diff --git "a/log/log-train-2023-02-05-17-58-35-0" "b/log/log-train-2023-02-05-17-58-35-0" new file mode 100644--- /dev/null +++ "b/log/log-train-2023-02-05-17-58-35-0" @@ -0,0 +1,25165 @@ +2023-02-05 17:58:35,365 INFO [train.py:973] (0/4) Training started +2023-02-05 17:58:35,371 INFO [train.py:983] (0/4) Device: cuda:0 +2023-02-05 17:58:35,412 INFO [train.py:992] (0/4) {'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] (0/4) About to create model +2023-02-05 17:58:36,048 INFO [zipformer.py:402] (0/4) 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 17:58:36,065 INFO [train.py:998] (0/4) Number of model parameters: 20697573 +2023-02-05 17:58:51,140 INFO [train.py:1013] (0/4) Using DDP +2023-02-05 17:58:51,426 INFO [asr_datamodule.py:420] (0/4) 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] (0/4) Enable MUSAN +2023-02-05 17:58:52,645 INFO [asr_datamodule.py:225] (0/4) About to get Musan cuts +2023-02-05 17:58:54,428 INFO [asr_datamodule.py:249] (0/4) Enable SpecAugment +2023-02-05 17:58:54,428 INFO [asr_datamodule.py:250] (0/4) Time warp factor: 80 +2023-02-05 17:58:54,428 INFO [asr_datamodule.py:260] (0/4) Num frame mask: 10 +2023-02-05 17:58:54,428 INFO [asr_datamodule.py:273] (0/4) About to create train dataset +2023-02-05 17:58:54,428 INFO [asr_datamodule.py:300] (0/4) Using DynamicBucketingSampler. +2023-02-05 17:58:54,448 WARNING [train.py:1067] (0/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-05 17:58:57,424 INFO [asr_datamodule.py:316] (0/4) About to create train dataloader +2023-02-05 17:58:57,424 INFO [asr_datamodule.py:430] (0/4) About to get dev-clean cuts +2023-02-05 17:58:57,425 INFO [asr_datamodule.py:437] (0/4) About to get dev-other cuts +2023-02-05 17:58:57,426 INFO [asr_datamodule.py:347] (0/4) About to create dev dataset +2023-02-05 17:58:57,789 INFO [asr_datamodule.py:364] (0/4) About to create dev dataloader +2023-02-05 17:59:07,110 WARNING [train.py:1067] (0/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-05 17:59:11,984 INFO [train.py:901] (0/4) Epoch 1, batch 0, loss[loss=7.062, simple_loss=6.39, pruned_loss=6.707, over 7704.00 frames. ], tot_loss[loss=7.062, simple_loss=6.39, pruned_loss=6.707, over 7704.00 frames. ], batch size: 18, lr: 2.50e-02, grad_scale: 2.0 +2023-02-05 17:59:11,985 INFO [train.py:926] (0/4) Computing validation loss +2023-02-05 17:59:24,177 INFO [train.py:935] (0/4) Epoch 1, validation: loss=6.888, simple_loss=6.229, pruned_loss=6.575, over 944034.00 frames. +2023-02-05 17:59:24,178 INFO [train.py:936] (0/4) Maximum memory allocated so far is 5748MB +2023-02-05 17:59:28,621 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=45.43 vs. limit=5.0 +2023-02-05 17:59:37,492 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=28.71 vs. limit=5.0 +2023-02-05 17:59:37,730 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-05 17:59:40,349 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=6.92 vs. limit=2.0 +2023-02-05 17:59:54,276 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=80.10 vs. limit=5.0 +2023-02-05 17:59:55,490 INFO [train.py:901] (0/4) Epoch 1, batch 50, loss[loss=1.445, simple_loss=1.28, pruned_loss=1.472, over 8763.00 frames. ], tot_loss[loss=2.165, simple_loss=1.956, pruned_loss=2.001, over 359569.46 frames. ], batch size: 30, lr: 2.75e-02, grad_scale: 0.25 +2023-02-05 17:59:56,133 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:00:06,234 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=22.09 vs. limit=2.0 +2023-02-05 18:00:11,291 WARNING [train.py:1067] (0/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-05 18:00:13,725 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:00:28,699 INFO [train.py:901] (0/4) Epoch 1, batch 100, loss[loss=1.19, simple_loss=1.018, pruned_loss=1.361, over 8101.00 frames. ], tot_loss[loss=1.649, simple_loss=1.468, pruned_loss=1.625, over 639424.52 frames. ], batch size: 23, lr: 3.00e-02, grad_scale: 0.0625 +2023-02-05 18:00:28,820 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=101.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:00:32,361 WARNING [train.py:1067] (0/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-05 18:00:32,814 INFO [optim.py:369] (0/4) 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 18:01:00,488 INFO [train.py:901] (0/4) Epoch 1, batch 150, loss[loss=1.051, simple_loss=0.8963, pruned_loss=1.124, over 8474.00 frames. ], tot_loss[loss=1.41, simple_loss=1.239, pruned_loss=1.441, over 856513.68 frames. ], batch size: 25, lr: 3.25e-02, grad_scale: 0.0625 +2023-02-05 18:01:02,597 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=40.01 vs. limit=5.0 +2023-02-05 18:01:18,442 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=86.14 vs. limit=5.0 +2023-02-05 18:01:27,981 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=3.24 vs. limit=2.0 +2023-02-05 18:01:34,596 INFO [train.py:901] (0/4) Epoch 1, batch 200, loss[loss=1, simple_loss=0.8467, pruned_loss=1.031, over 8478.00 frames. ], tot_loss[loss=1.27, simple_loss=1.106, pruned_loss=1.306, over 1025749.67 frames. ], batch size: 25, lr: 3.50e-02, grad_scale: 0.125 +2023-02-05 18:01:37,992 INFO [optim.py:369] (0/4) 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 18:01:45,609 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=2.52 vs. limit=2.0 +2023-02-05 18:01:50,414 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=17.46 vs. limit=5.0 +2023-02-05 18:02:05,438 INFO [train.py:901] (0/4) Epoch 1, batch 250, loss[loss=0.8579, simple_loss=0.7197, pruned_loss=0.861, over 7421.00 frames. ], tot_loss[loss=1.182, simple_loss=1.021, pruned_loss=1.209, over 1155566.62 frames. ], batch size: 17, lr: 3.75e-02, grad_scale: 0.125 +2023-02-05 18:02:14,823 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-05 18:02:22,944 WARNING [train.py:1067] (0/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-05 18:02:23,769 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=11.14 vs. limit=5.0 +2023-02-05 18:02:37,913 INFO [train.py:901] (0/4) Epoch 1, batch 300, loss[loss=0.9402, simple_loss=0.7819, pruned_loss=0.9226, over 8249.00 frames. ], tot_loss[loss=1.132, simple_loss=0.9702, pruned_loss=1.146, over 1266587.67 frames. ], batch size: 24, lr: 4.00e-02, grad_scale: 0.25 +2023-02-05 18:02:42,325 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=306.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:02:42,690 INFO [optim.py:369] (0/4) 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 18:02:46,523 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=2.09 vs. limit=2.0 +2023-02-05 18:02:47,403 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=314.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:03:10,258 INFO [train.py:901] (0/4) Epoch 1, batch 350, loss[loss=0.9084, simple_loss=0.7508, pruned_loss=0.8663, over 7651.00 frames. ], tot_loss[loss=1.089, simple_loss=0.9264, pruned_loss=1.089, over 1341372.47 frames. ], batch size: 19, lr: 4.25e-02, grad_scale: 0.25 +2023-02-05 18:03:30,777 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=5.41 vs. limit=5.0 +2023-02-05 18:03:42,314 INFO [train.py:901] (0/4) Epoch 1, batch 400, loss[loss=0.9906, simple_loss=0.8123, pruned_loss=0.9259, over 8327.00 frames. ], tot_loss[loss=1.061, simple_loss=0.8956, pruned_loss=1.047, over 1406491.97 frames. ], batch size: 26, lr: 4.50e-02, grad_scale: 0.5 +2023-02-05 18:03:44,609 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=405.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:03:45,466 INFO [optim.py:369] (0/4) 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 18:03:55,280 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=421.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:04:11,512 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=445.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:04:15,517 INFO [train.py:901] (0/4) Epoch 1, batch 450, loss[loss=0.9387, simple_loss=0.766, pruned_loss=0.8553, over 8087.00 frames. ], tot_loss[loss=1.035, simple_loss=0.8675, pruned_loss=1.006, over 1448980.14 frames. ], batch size: 21, lr: 4.75e-02, grad_scale: 0.5 +2023-02-05 18:04:18,234 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.6355, 5.6725, 5.6731, 5.6728, 5.6732, 5.6732, 5.6708, 5.6731], + device='cuda:0'), covar=tensor([0.0025, 0.0032, 0.0029, 0.0036, 0.0031, 0.0039, 0.0033, 0.0036], + device='cuda:0'), in_proj_covar=tensor([0.0012, 0.0013, 0.0013, 0.0014, 0.0012, 0.0013, 0.0014, 0.0013], + device='cuda:0'), out_proj_covar=tensor([8.7666e-06, 9.0568e-06, 9.0095e-06, 8.9489e-06, 9.1031e-06, 8.8468e-06, + 8.8892e-06, 8.9351e-06], device='cuda:0') +2023-02-05 18:04:25,496 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.3666, 2.3279, 4.4231, 4.4585, 4.5501, 3.8236, 3.8460, 4.3119], + device='cuda:0'), covar=tensor([0.0142, 0.0450, 0.0182, 0.0136, 0.0102, 0.0175, 0.0308, 0.0158], + device='cuda:0'), in_proj_covar=tensor([0.0014, 0.0015, 0.0015, 0.0015, 0.0015, 0.0015, 0.0016, 0.0014], + device='cuda:0'), out_proj_covar=tensor([9.2644e-06, 9.7546e-06, 9.5960e-06, 9.5450e-06, 9.3455e-06, 9.5027e-06, + 9.6717e-06, 9.3750e-06], device='cuda:0') +2023-02-05 18:04:45,729 INFO [train.py:901] (0/4) Epoch 1, batch 500, loss[loss=1.024, simple_loss=0.8321, pruned_loss=0.9099, over 8239.00 frames. ], tot_loss[loss=1.016, simple_loss=0.8458, pruned_loss=0.9703, over 1484599.60 frames. ], batch size: 24, lr: 4.99e-02, grad_scale: 1.0 +2023-02-05 18:04:49,472 INFO [optim.py:369] (0/4) 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 18:05:04,099 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=10.81 vs. limit=5.0 +2023-02-05 18:05:16,930 INFO [train.py:901] (0/4) Epoch 1, batch 550, loss[loss=0.8317, simple_loss=0.6842, pruned_loss=0.6948, over 7420.00 frames. ], tot_loss[loss=1, simple_loss=0.8295, pruned_loss=0.9346, over 1512549.36 frames. ], batch size: 17, lr: 4.98e-02, grad_scale: 1.0 +2023-02-05 18:05:22,218 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=560.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:05:33,864 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=580.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:05:39,251 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=586.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:05:47,854 INFO [train.py:901] (0/4) Epoch 1, batch 600, loss[loss=0.933, simple_loss=0.774, pruned_loss=0.7431, over 8364.00 frames. ], tot_loss[loss=0.9849, simple_loss=0.8158, pruned_loss=0.8952, over 1534391.37 frames. ], batch size: 26, lr: 4.98e-02, grad_scale: 1.0 +2023-02-05 18:05:51,149 INFO [optim.py:369] (0/4) 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 18:05:51,940 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=608.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:05:57,480 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-05 18:06:15,545 INFO [train.py:901] (0/4) Epoch 1, batch 650, loss[loss=0.7697, simple_loss=0.6433, pruned_loss=0.5881, over 7413.00 frames. ], tot_loss[loss=0.9651, simple_loss=0.8001, pruned_loss=0.8506, over 1553972.71 frames. ], batch size: 17, lr: 4.98e-02, grad_scale: 1.0 +2023-02-05 18:06:16,667 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.6573, 3.2249, 4.2544, 3.3797, 3.9081, 4.0795, 4.1592, 4.2451], + device='cuda:0'), covar=tensor([0.3645, 0.1044, 0.0146, 0.1674, 0.1096, 0.0480, 0.0314, 0.0163], + device='cuda:0'), in_proj_covar=tensor([0.0015, 0.0015, 0.0015, 0.0015, 0.0014, 0.0014, 0.0014, 0.0014], + device='cuda:0'), out_proj_covar=tensor([1.0089e-05, 1.0593e-05, 9.6553e-06, 1.0564e-05, 9.8548e-06, 9.6059e-06, + 9.4800e-06, 9.3276e-06], device='cuda:0') +2023-02-05 18:06:18,948 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0317, 0.9921, 1.2374, 1.4060, 1.0335, 1.0047, 1.2737, 1.1100], + device='cuda:0'), covar=tensor([0.6608, 0.9232, 0.5610, 0.3609, 0.4855, 0.6556, 0.4504, 0.5955], + device='cuda:0'), in_proj_covar=tensor([0.0043, 0.0048, 0.0040, 0.0036, 0.0039, 0.0050, 0.0039, 0.0043], + device='cuda:0'), out_proj_covar=tensor([2.8100e-05, 3.2251e-05, 2.9394e-05, 2.2395e-05, 2.5689e-05, 2.9535e-05, + 2.6128e-05, 2.7634e-05], device='cuda:0') +2023-02-05 18:06:20,640 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=658.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:06:31,062 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=677.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:06:44,419 INFO [train.py:901] (0/4) Epoch 1, batch 700, loss[loss=0.8819, simple_loss=0.7296, pruned_loss=0.6744, over 8236.00 frames. ], tot_loss[loss=0.9394, simple_loss=0.7808, pruned_loss=0.8021, over 1567608.44 frames. ], batch size: 22, lr: 4.98e-02, grad_scale: 1.0 +2023-02-05 18:06:45,058 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=702.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:06:48,203 INFO [optim.py:369] (0/4) 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 18:07:14,466 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=749.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:07:15,374 INFO [train.py:901] (0/4) Epoch 1, batch 750, loss[loss=0.6826, simple_loss=0.5745, pruned_loss=0.493, over 7263.00 frames. ], tot_loss[loss=0.9109, simple_loss=0.7595, pruned_loss=0.7538, over 1579447.87 frames. ], batch size: 16, lr: 4.97e-02, grad_scale: 1.0 +2023-02-05 18:07:25,628 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-05 18:07:26,836 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=773.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:07:32,315 WARNING [train.py:1067] (0/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-05 18:07:43,634 INFO [train.py:901] (0/4) Epoch 1, batch 800, loss[loss=0.7957, simple_loss=0.6825, pruned_loss=0.5425, over 8486.00 frames. ], tot_loss[loss=0.8849, simple_loss=0.7407, pruned_loss=0.7095, over 1589739.26 frames. ], batch size: 28, lr: 4.97e-02, grad_scale: 2.0 +2023-02-05 18:07:46,609 INFO [optim.py:369] (0/4) 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 18:07:46,898 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=2.03 vs. limit=2.0 +2023-02-05 18:07:51,294 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=816.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:08:05,152 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=841.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:08:11,179 INFO [train.py:901] (0/4) Epoch 1, batch 850, loss[loss=0.8285, simple_loss=0.7108, pruned_loss=0.5569, over 8454.00 frames. ], tot_loss[loss=0.8578, simple_loss=0.7211, pruned_loss=0.6672, over 1597597.96 frames. ], batch size: 27, lr: 4.96e-02, grad_scale: 2.0 +2023-02-05 18:08:22,413 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=864.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:08:22,881 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=865.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:08:42,859 INFO [train.py:901] (0/4) Epoch 1, batch 900, loss[loss=0.6883, simple_loss=0.5921, pruned_loss=0.4541, over 7689.00 frames. ], tot_loss[loss=0.8311, simple_loss=0.7016, pruned_loss=0.628, over 1599772.90 frames. ], batch size: 18, lr: 4.96e-02, grad_scale: 2.0 +2023-02-05 18:08:46,413 INFO [optim.py:369] (0/4) 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 18:08:55,002 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.1197, 0.7853, 1.2201, 1.6118, 1.0989, 1.0041, 1.3250, 1.3545], + device='cuda:0'), covar=tensor([1.7977, 2.5836, 1.4440, 0.9846, 1.5354, 1.5254, 1.7958, 1.9836], + device='cuda:0'), in_proj_covar=tensor([0.0090, 0.0087, 0.0081, 0.0075, 0.0081, 0.0089, 0.0091, 0.0095], + device='cuda:0'), out_proj_covar=tensor([5.9339e-05, 6.2556e-05, 5.7150e-05, 4.4411e-05, 5.6099e-05, 5.6563e-05, + 6.1271e-05, 6.3668e-05], device='cuda:0') +2023-02-05 18:08:55,935 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=924.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:08:58,993 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=930.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:09:10,100 INFO [train.py:901] (0/4) Epoch 1, batch 950, loss[loss=0.7954, simple_loss=0.6831, pruned_loss=0.5202, over 8488.00 frames. ], tot_loss[loss=0.8075, simple_loss=0.6847, pruned_loss=0.5933, over 1600496.14 frames. ], batch size: 29, lr: 4.96e-02, grad_scale: 2.0 +2023-02-05 18:09:10,749 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=952.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:09:16,849 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=2.15 vs. limit=2.0 +2023-02-05 18:09:18,365 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.78 vs. limit=2.0 +2023-02-05 18:09:26,439 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-05 18:09:37,675 INFO [train.py:901] (0/4) Epoch 1, batch 1000, loss[loss=0.7615, simple_loss=0.6534, pruned_loss=0.4934, over 8560.00 frames. ], tot_loss[loss=0.7857, simple_loss=0.6689, pruned_loss=0.5628, over 1600277.08 frames. ], batch size: 31, lr: 4.95e-02, grad_scale: 2.0 +2023-02-05 18:09:40,947 INFO [optim.py:369] (0/4) 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 18:09:49,495 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.82 vs. limit=2.0 +2023-02-05 18:09:52,900 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1029.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:09:53,918 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-05 18:09:59,173 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1039.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:09:59,970 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 +2023-02-05 18:10:02,610 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1045.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:10:05,083 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-05 18:10:05,584 INFO [train.py:901] (0/4) Epoch 1, batch 1050, loss[loss=0.7116, simple_loss=0.6135, pruned_loss=0.4525, over 8238.00 frames. ], tot_loss[loss=0.7675, simple_loss=0.6559, pruned_loss=0.5365, over 1602544.88 frames. ], batch size: 22, lr: 4.95e-02, grad_scale: 2.0 +2023-02-05 18:10:07,185 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1054.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:10:14,060 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1067.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:10:33,037 INFO [train.py:901] (0/4) Epoch 1, batch 1100, loss[loss=0.7241, simple_loss=0.6334, pruned_loss=0.4451, over 8350.00 frames. ], tot_loss[loss=0.7482, simple_loss=0.6421, pruned_loss=0.511, over 1607440.84 frames. ], batch size: 24, lr: 4.94e-02, grad_scale: 2.0 +2023-02-05 18:10:36,085 INFO [optim.py:369] (0/4) 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 18:10:43,721 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1120.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:10:56,853 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1145.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:10:59,912 INFO [train.py:901] (0/4) Epoch 1, batch 1150, loss[loss=0.6159, simple_loss=0.5411, pruned_loss=0.3729, over 8025.00 frames. ], tot_loss[loss=0.7292, simple_loss=0.6287, pruned_loss=0.4871, over 1608944.87 frames. ], batch size: 22, lr: 4.94e-02, grad_scale: 2.0 +2023-02-05 18:11:01,596 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-05 18:11:11,751 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1171.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:11:27,730 INFO [train.py:901] (0/4) Epoch 1, batch 1200, loss[loss=0.7309, simple_loss=0.6416, pruned_loss=0.4403, over 8489.00 frames. ], tot_loss[loss=0.7181, simple_loss=0.6218, pruned_loss=0.4697, over 1610304.18 frames. ], batch size: 29, lr: 4.93e-02, grad_scale: 4.0 +2023-02-05 18:11:30,970 INFO [optim.py:369] (0/4) 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 18:11:32,131 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=1209.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:11:56,781 INFO [train.py:901] (0/4) Epoch 1, batch 1250, loss[loss=0.6768, simple_loss=0.5954, pruned_loss=0.4037, over 8067.00 frames. ], tot_loss[loss=0.7047, simple_loss=0.6128, pruned_loss=0.4522, over 1614248.95 frames. ], batch size: 21, lr: 4.92e-02, grad_scale: 4.0 +2023-02-05 18:12:21,152 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1295.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:12:24,264 INFO [train.py:901] (0/4) Epoch 1, batch 1300, loss[loss=0.6242, simple_loss=0.5567, pruned_loss=0.3627, over 8128.00 frames. ], tot_loss[loss=0.6926, simple_loss=0.6043, pruned_loss=0.437, over 1613115.47 frames. ], batch size: 22, lr: 4.92e-02, grad_scale: 4.0 +2023-02-05 18:12:24,443 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1301.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:12:27,427 INFO [optim.py:369] (0/4) 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 18:12:34,702 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1320.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:12:36,281 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1323.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:12:36,743 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1324.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:12:37,927 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1326.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:12:51,928 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1348.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:12:53,375 INFO [train.py:901] (0/4) Epoch 1, batch 1350, loss[loss=0.6582, simple_loss=0.5873, pruned_loss=0.3804, over 8238.00 frames. ], tot_loss[loss=0.6793, simple_loss=0.5948, pruned_loss=0.4219, over 1611014.27 frames. ], batch size: 22, lr: 4.91e-02, grad_scale: 4.0 +2023-02-05 18:13:22,443 INFO [train.py:901] (0/4) Epoch 1, batch 1400, loss[loss=0.5712, simple_loss=0.5087, pruned_loss=0.3294, over 7676.00 frames. ], tot_loss[loss=0.6715, simple_loss=0.5899, pruned_loss=0.4111, over 1618952.52 frames. ], batch size: 18, lr: 4.91e-02, grad_scale: 4.0 +2023-02-05 18:13:25,824 INFO [optim.py:369] (0/4) 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 18:13:39,765 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([0.9694, 0.9354, 1.0524, 1.2808, 0.8859, 0.7541, 0.7290, 1.2804], + device='cuda:0'), covar=tensor([0.9348, 0.9147, 0.7403, 0.3867, 0.9678, 1.0891, 0.9906, 0.7617], + device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0132, 0.0124, 0.0102, 0.0155, 0.0151, 0.0146, 0.0142], + device='cuda:0'), out_proj_covar=tensor([9.5277e-05, 9.4346e-05, 8.9994e-05, 6.2316e-05, 1.0979e-04, 1.0418e-04, + 1.0368e-04, 9.9812e-05], device='cuda:0') +2023-02-05 18:13:50,937 INFO [train.py:901] (0/4) Epoch 1, batch 1450, loss[loss=0.5333, simple_loss=0.5039, pruned_loss=0.2818, over 8094.00 frames. ], tot_loss[loss=0.6617, simple_loss=0.5835, pruned_loss=0.3995, over 1615644.80 frames. ], batch size: 21, lr: 4.90e-02, grad_scale: 4.0 +2023-02-05 18:13:51,607 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1452.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:13:54,972 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-05 18:14:21,306 INFO [train.py:901] (0/4) Epoch 1, batch 1500, loss[loss=0.6424, simple_loss=0.5767, pruned_loss=0.3638, over 8516.00 frames. ], tot_loss[loss=0.6531, simple_loss=0.5784, pruned_loss=0.3891, over 1621508.33 frames. ], batch size: 26, lr: 4.89e-02, grad_scale: 4.0 +2023-02-05 18:14:24,736 INFO [optim.py:369] (0/4) 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] (0/4) 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] (0/4) Epoch 1, batch 1550, loss[loss=0.6685, simple_loss=0.5763, pruned_loss=0.3956, over 7576.00 frames. ], tot_loss[loss=0.6469, simple_loss=0.5743, pruned_loss=0.3814, over 1620758.52 frames. ], batch size: 73, lr: 4.89e-02, grad_scale: 4.0 +2023-02-05 18:14:54,640 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.96 vs. limit=2.0 +2023-02-05 18:15:08,618 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1580.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:15:10,869 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1584.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:15:20,753 INFO [train.py:901] (0/4) Epoch 1, batch 1600, loss[loss=0.5118, simple_loss=0.4812, pruned_loss=0.272, over 7547.00 frames. ], tot_loss[loss=0.6388, simple_loss=0.5685, pruned_loss=0.3731, over 1621122.73 frames. ], batch size: 18, lr: 4.88e-02, grad_scale: 8.0 +2023-02-05 18:15:23,969 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1605.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:15:24,967 INFO [optim.py:369] (0/4) 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 18:15:32,188 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.95 vs. limit=2.0 +2023-02-05 18:15:37,780 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1629.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:15:38,281 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1630.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:15:50,679 INFO [train.py:901] (0/4) Epoch 1, batch 1650, loss[loss=0.6551, simple_loss=0.5707, pruned_loss=0.3795, over 6814.00 frames. ], tot_loss[loss=0.6306, simple_loss=0.5626, pruned_loss=0.365, over 1614061.40 frames. ], batch size: 71, lr: 4.87e-02, grad_scale: 8.0 +2023-02-05 18:16:21,954 INFO [train.py:901] (0/4) Epoch 1, batch 1700, loss[loss=0.5567, simple_loss=0.5041, pruned_loss=0.3084, over 7698.00 frames. ], tot_loss[loss=0.6194, simple_loss=0.5553, pruned_loss=0.3548, over 1609010.73 frames. ], batch size: 18, lr: 4.86e-02, grad_scale: 8.0 +2023-02-05 18:16:25,350 INFO [optim.py:369] (0/4) 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 18:16:49,137 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4064, 2.5674, 1.5729, 2.3622, 2.3275, 2.1960, 2.0051, 2.7086], + device='cuda:0'), covar=tensor([0.3527, 0.3549, 0.4976, 0.3033, 0.4125, 0.4216, 0.3629, 0.2976], + device='cuda:0'), in_proj_covar=tensor([0.0118, 0.0108, 0.0094, 0.0100, 0.0136, 0.0113, 0.0099, 0.0114], + device='cuda:0'), out_proj_covar=tensor([8.9427e-05, 7.7024e-05, 7.0985e-05, 7.5784e-05, 9.8553e-05, 8.2951e-05, + 7.6344e-05, 8.3422e-05], device='cuda:0') +2023-02-05 18:16:51,249 INFO [train.py:901] (0/4) Epoch 1, batch 1750, loss[loss=0.6061, simple_loss=0.5627, pruned_loss=0.3264, over 8558.00 frames. ], tot_loss[loss=0.6146, simple_loss=0.5525, pruned_loss=0.3493, over 1613329.64 frames. ], batch size: 31, lr: 4.86e-02, grad_scale: 8.0 +2023-02-05 18:17:18,058 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=1796.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:17:21,113 INFO [train.py:901] (0/4) Epoch 1, batch 1800, loss[loss=0.5207, simple_loss=0.479, pruned_loss=0.2827, over 7800.00 frames. ], tot_loss[loss=0.6089, simple_loss=0.5491, pruned_loss=0.3434, over 1610477.81 frames. ], batch size: 19, lr: 4.85e-02, grad_scale: 8.0 +2023-02-05 18:17:24,722 INFO [optim.py:369] (0/4) 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 18:17:52,117 INFO [train.py:901] (0/4) Epoch 1, batch 1850, loss[loss=0.5165, simple_loss=0.4981, pruned_loss=0.2669, over 8496.00 frames. ], tot_loss[loss=0.6049, simple_loss=0.5472, pruned_loss=0.3387, over 1617370.65 frames. ], batch size: 26, lr: 4.84e-02, grad_scale: 8.0 +2023-02-05 18:17:55,051 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1856.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:18:06,750 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1875.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:18:13,284 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1886.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:18:14,328 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 1, batch 1900, loss[loss=0.5806, simple_loss=0.5413, pruned_loss=0.3104, over 8046.00 frames. ], tot_loss[loss=0.5965, simple_loss=0.5426, pruned_loss=0.3312, over 1616765.15 frames. ], batch size: 22, lr: 4.83e-02, grad_scale: 8.0 +2023-02-05 18:18:25,476 INFO [optim.py:369] (0/4) 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 18:18:27,923 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1911.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:18:27,934 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1911.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:18:37,725 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=1928.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:18:45,004 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-05 18:18:47,529 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.2294, 1.7132, 3.1471, 2.2438, 3.0025, 2.8859, 2.8379, 2.8852], + device='cuda:0'), covar=tensor([0.0199, 0.1875, 0.0246, 0.0525, 0.0236, 0.0229, 0.0289, 0.0322], + device='cuda:0'), in_proj_covar=tensor([0.0053, 0.0148, 0.0070, 0.0088, 0.0071, 0.0070, 0.0079, 0.0088], + device='cuda:0'), out_proj_covar=tensor([3.1692e-05, 9.3882e-05, 4.2070e-05, 5.8846e-05, 4.0672e-05, 3.9936e-05, + 4.6738e-05, 5.2428e-05], device='cuda:0') +2023-02-05 18:18:52,617 INFO [train.py:901] (0/4) Epoch 1, batch 1950, loss[loss=0.6157, simple_loss=0.5609, pruned_loss=0.3358, over 8607.00 frames. ], tot_loss[loss=0.5902, simple_loss=0.5384, pruned_loss=0.3257, over 1614127.86 frames. ], batch size: 31, lr: 4.83e-02, grad_scale: 8.0 +2023-02-05 18:18:55,546 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-05 18:19:05,701 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=1973.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:19:11,330 WARNING [train.py:1067] (0/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-05 18:19:13,569 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 +2023-02-05 18:19:22,194 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-2000.pt +2023-02-05 18:19:23,718 INFO [train.py:901] (0/4) Epoch 1, batch 2000, loss[loss=0.5396, simple_loss=0.5064, pruned_loss=0.2864, over 8025.00 frames. ], tot_loss[loss=0.5858, simple_loss=0.5364, pruned_loss=0.3213, over 1617582.26 frames. ], batch size: 22, lr: 4.82e-02, grad_scale: 8.0 +2023-02-05 18:19:27,549 INFO [optim.py:369] (0/4) 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 18:19:50,353 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2043.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:19:56,664 INFO [train.py:901] (0/4) Epoch 1, batch 2050, loss[loss=0.6043, simple_loss=0.5602, pruned_loss=0.3242, over 8343.00 frames. ], tot_loss[loss=0.5768, simple_loss=0.5311, pruned_loss=0.3141, over 1620229.34 frames. ], batch size: 26, lr: 4.81e-02, grad_scale: 8.0 +2023-02-05 18:20:03,236 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7805, 1.8276, 1.8462, 1.9054, 1.5613, 1.7554, 0.6695, 1.0874], + device='cuda:0'), covar=tensor([0.1339, 0.0623, 0.0811, 0.0911, 0.1394, 0.0983, 0.2971, 0.1474], + device='cuda:0'), in_proj_covar=tensor([0.0090, 0.0073, 0.0077, 0.0083, 0.0098, 0.0074, 0.0120, 0.0094], + device='cuda:0'), out_proj_covar=tensor([6.1794e-05, 4.9907e-05, 5.0106e-05, 5.7602e-05, 7.0531e-05, 4.6409e-05, + 8.3969e-05, 6.6210e-05], device='cuda:0') +2023-02-05 18:20:21,043 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2088.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:20:29,073 INFO [train.py:901] (0/4) Epoch 1, batch 2100, loss[loss=0.5244, simple_loss=0.4981, pruned_loss=0.2754, over 7913.00 frames. ], tot_loss[loss=0.5691, simple_loss=0.5268, pruned_loss=0.308, over 1620915.09 frames. ], batch size: 20, lr: 4.80e-02, grad_scale: 16.0 +2023-02-05 18:20:32,718 INFO [optim.py:369] (0/4) 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 18:21:01,649 INFO [train.py:901] (0/4) Epoch 1, batch 2150, loss[loss=0.5392, simple_loss=0.4938, pruned_loss=0.2923, over 7541.00 frames. ], tot_loss[loss=0.5616, simple_loss=0.5232, pruned_loss=0.3017, over 1623217.31 frames. ], batch size: 18, lr: 4.79e-02, grad_scale: 16.0 +2023-02-05 18:21:11,749 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2167.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:21:29,901 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2192.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:21:35,019 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=2200.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:21:35,571 INFO [train.py:901] (0/4) Epoch 1, batch 2200, loss[loss=0.4796, simple_loss=0.4641, pruned_loss=0.2476, over 7821.00 frames. ], tot_loss[loss=0.5524, simple_loss=0.5178, pruned_loss=0.2949, over 1622082.51 frames. ], batch size: 20, lr: 4.78e-02, grad_scale: 16.0 +2023-02-05 18:21:39,334 INFO [optim.py:369] (0/4) 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 18:21:40,788 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.5004, 1.3296, 4.3680, 2.4653, 4.0451, 3.7920, 3.6528, 3.7748], + device='cuda:0'), covar=tensor([0.0179, 0.3444, 0.0191, 0.1034, 0.0279, 0.0231, 0.0346, 0.0371], + device='cuda:0'), in_proj_covar=tensor([0.0061, 0.0174, 0.0077, 0.0097, 0.0083, 0.0079, 0.0085, 0.0096], + device='cuda:0'), out_proj_covar=tensor([3.8018e-05, 1.0834e-04, 4.7521e-05, 6.5599e-05, 4.6665e-05, 4.3482e-05, + 4.9092e-05, 5.6088e-05], device='cuda:0') +2023-02-05 18:21:46,984 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=2219.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:21:55,781 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=2232.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:22:07,870 INFO [train.py:901] (0/4) Epoch 1, batch 2250, loss[loss=0.6515, simple_loss=0.583, pruned_loss=0.36, over 7097.00 frames. ], tot_loss[loss=0.5446, simple_loss=0.5132, pruned_loss=0.2891, over 1617345.32 frames. ], batch size: 72, lr: 4.77e-02, grad_scale: 16.0 +2023-02-05 18:22:41,023 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2299.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:22:42,096 INFO [train.py:901] (0/4) Epoch 1, batch 2300, loss[loss=0.4982, simple_loss=0.473, pruned_loss=0.2617, over 7551.00 frames. ], tot_loss[loss=0.5405, simple_loss=0.511, pruned_loss=0.2858, over 1618424.39 frames. ], batch size: 18, lr: 4.77e-02, grad_scale: 16.0 +2023-02-05 18:22:45,954 INFO [optim.py:369] (0/4) 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 18:22:51,203 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2315.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:22:56,953 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2324.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:23:03,184 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2334.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:23:09,675 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2344.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:23:12,274 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2347.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:23:14,709 INFO [train.py:901] (0/4) Epoch 1, batch 2350, loss[loss=0.4615, simple_loss=0.4622, pruned_loss=0.2304, over 8029.00 frames. ], tot_loss[loss=0.5343, simple_loss=0.5075, pruned_loss=0.2812, over 1613236.33 frames. ], batch size: 22, lr: 4.76e-02, grad_scale: 16.0 +2023-02-05 18:23:19,261 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2358.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:23:26,044 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2369.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:23:41,127 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 +2023-02-05 18:23:46,436 INFO [train.py:901] (0/4) Epoch 1, batch 2400, loss[loss=0.4752, simple_loss=0.4801, pruned_loss=0.2352, over 8324.00 frames. ], tot_loss[loss=0.5297, simple_loss=0.5057, pruned_loss=0.2774, over 1618649.05 frames. ], batch size: 26, lr: 4.75e-02, grad_scale: 16.0 +2023-02-05 18:23:50,349 INFO [optim.py:369] (0/4) 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 18:24:13,820 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.1515, 1.1506, 1.0279, 1.1471, 0.8239, 0.8154, 0.9169, 1.1843], + device='cuda:0'), covar=tensor([0.2211, 0.1995, 0.2143, 0.0939, 0.3205, 0.3088, 0.2719, 0.2139], + device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0190, 0.0170, 0.0135, 0.0242, 0.0215, 0.0241, 0.0199], + device='cuda:0'), out_proj_covar=tensor([1.4572e-04, 1.4395e-04, 1.3575e-04, 9.5700e-05, 1.7650e-04, 1.5924e-04, + 1.7827e-04, 1.5342e-04], device='cuda:0') +2023-02-05 18:24:15,213 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3516, 1.9242, 2.9591, 1.4109, 2.0315, 2.0076, 1.2200, 2.0083], + device='cuda:0'), covar=tensor([0.2459, 0.2253, 0.0331, 0.1819, 0.1667, 0.2409, 0.2489, 0.1920], + device='cuda:0'), in_proj_covar=tensor([0.0105, 0.0110, 0.0060, 0.0092, 0.0127, 0.0124, 0.0110, 0.0133], + device='cuda:0'), out_proj_covar=tensor([7.3969e-05, 7.5969e-05, 3.8170e-05, 6.1394e-05, 8.3709e-05, 9.0299e-05, + 7.1550e-05, 8.9489e-05], device='cuda:0') +2023-02-05 18:24:20,800 INFO [train.py:901] (0/4) Epoch 1, batch 2450, loss[loss=0.5398, simple_loss=0.5212, pruned_loss=0.2792, over 8456.00 frames. ], tot_loss[loss=0.5269, simple_loss=0.5035, pruned_loss=0.2756, over 1613909.64 frames. ], batch size: 29, lr: 4.74e-02, grad_scale: 16.0 +2023-02-05 18:24:52,759 INFO [train.py:901] (0/4) Epoch 1, batch 2500, loss[loss=0.5914, simple_loss=0.5514, pruned_loss=0.3157, over 8488.00 frames. ], tot_loss[loss=0.5255, simple_loss=0.5032, pruned_loss=0.2742, over 1616523.58 frames. ], batch size: 28, lr: 4.73e-02, grad_scale: 16.0 +2023-02-05 18:24:56,549 INFO [optim.py:369] (0/4) 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 18:25:19,774 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=2.01 vs. limit=2.0 +2023-02-05 18:25:25,600 INFO [train.py:901] (0/4) Epoch 1, batch 2550, loss[loss=0.5509, simple_loss=0.5325, pruned_loss=0.2847, over 8475.00 frames. ], tot_loss[loss=0.5264, simple_loss=0.504, pruned_loss=0.2747, over 1618999.91 frames. ], batch size: 25, lr: 4.72e-02, grad_scale: 16.0 +2023-02-05 18:25:38,474 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2571.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:25:51,084 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2590.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:25:54,852 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2596.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:25:57,891 INFO [train.py:901] (0/4) Epoch 1, batch 2600, loss[loss=0.5284, simple_loss=0.5172, pruned_loss=0.2698, over 8244.00 frames. ], tot_loss[loss=0.5194, simple_loss=0.4999, pruned_loss=0.2696, over 1616207.34 frames. ], batch size: 22, lr: 4.71e-02, grad_scale: 16.0 +2023-02-05 18:25:59,371 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2603.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:26:01,603 INFO [optim.py:369] (0/4) 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 18:26:06,861 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2615.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:26:08,422 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 +2023-02-05 18:26:15,228 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2628.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:26:31,161 INFO [train.py:901] (0/4) Epoch 1, batch 2650, loss[loss=0.4195, simple_loss=0.4358, pruned_loss=0.2016, over 8355.00 frames. ], tot_loss[loss=0.5147, simple_loss=0.4978, pruned_loss=0.2659, over 1613494.11 frames. ], batch size: 24, lr: 4.70e-02, grad_scale: 16.0 +2023-02-05 18:26:37,067 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5339, 1.9879, 4.0508, 1.5052, 2.5483, 2.4491, 1.4861, 2.4751], + device='cuda:0'), covar=tensor([0.2093, 0.2260, 0.0176, 0.1712, 0.1719, 0.2300, 0.2137, 0.1856], + device='cuda:0'), in_proj_covar=tensor([0.0111, 0.0112, 0.0065, 0.0101, 0.0137, 0.0140, 0.0118, 0.0149], + device='cuda:0'), out_proj_covar=tensor([7.9182e-05, 7.7887e-05, 4.0440e-05, 6.8655e-05, 9.1238e-05, 1.0226e-04, + 7.7374e-05, 9.9452e-05], device='cuda:0') +2023-02-05 18:27:03,817 INFO [train.py:901] (0/4) Epoch 1, batch 2700, loss[loss=0.47, simple_loss=0.48, pruned_loss=0.23, over 8345.00 frames. ], tot_loss[loss=0.5117, simple_loss=0.4954, pruned_loss=0.2641, over 1616729.86 frames. ], batch size: 26, lr: 4.69e-02, grad_scale: 16.0 +2023-02-05 18:27:04,575 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=2702.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:27:05,219 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2703.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:27:08,312 INFO [optim.py:369] (0/4) 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 18:27:37,287 INFO [train.py:901] (0/4) Epoch 1, batch 2750, loss[loss=0.489, simple_loss=0.4891, pruned_loss=0.2445, over 8025.00 frames. ], tot_loss[loss=0.5066, simple_loss=0.4935, pruned_loss=0.26, over 1620489.35 frames. ], batch size: 22, lr: 4.68e-02, grad_scale: 16.0 +2023-02-05 18:28:05,773 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5588, 1.7939, 2.2013, 1.5346, 1.6040, 1.9216, 1.0421, 1.3819], + device='cuda:0'), covar=tensor([0.1407, 0.0615, 0.0449, 0.0917, 0.0879, 0.1040, 0.1960, 0.1275], + device='cuda:0'), in_proj_covar=tensor([0.0108, 0.0089, 0.0081, 0.0090, 0.0098, 0.0084, 0.0140, 0.0124], + device='cuda:0'), out_proj_covar=tensor([7.6570e-05, 6.3812e-05, 5.5964e-05, 6.5291e-05, 7.0947e-05, 5.8194e-05, + 1.0517e-04, 9.4265e-05], device='cuda:0') +2023-02-05 18:28:11,566 INFO [train.py:901] (0/4) Epoch 1, batch 2800, loss[loss=0.4946, simple_loss=0.4879, pruned_loss=0.2507, over 8026.00 frames. ], tot_loss[loss=0.502, simple_loss=0.4904, pruned_loss=0.2569, over 1614532.32 frames. ], batch size: 22, lr: 4.67e-02, grad_scale: 16.0 +2023-02-05 18:28:15,257 INFO [optim.py:369] (0/4) 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 18:28:21,889 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2817.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:28:38,508 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2842.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:28:44,178 INFO [train.py:901] (0/4) Epoch 1, batch 2850, loss[loss=0.4538, simple_loss=0.4536, pruned_loss=0.227, over 7976.00 frames. ], tot_loss[loss=0.5012, simple_loss=0.4899, pruned_loss=0.2563, over 1611591.54 frames. ], batch size: 21, lr: 4.66e-02, grad_scale: 16.0 +2023-02-05 18:29:07,282 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=2.01 vs. limit=2.0 +2023-02-05 18:29:18,794 INFO [train.py:901] (0/4) Epoch 1, batch 2900, loss[loss=0.4998, simple_loss=0.4958, pruned_loss=0.2519, over 8589.00 frames. ], tot_loss[loss=0.5007, simple_loss=0.4898, pruned_loss=0.2558, over 1612376.32 frames. ], batch size: 31, lr: 4.65e-02, grad_scale: 16.0 +2023-02-05 18:29:22,677 INFO [optim.py:369] (0/4) 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 18:29:38,018 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=2.26 vs. limit=2.0 +2023-02-05 18:29:48,934 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-05 18:29:52,162 INFO [train.py:901] (0/4) Epoch 1, batch 2950, loss[loss=0.4712, simple_loss=0.4682, pruned_loss=0.2371, over 7941.00 frames. ], tot_loss[loss=0.4998, simple_loss=0.4899, pruned_loss=0.2549, over 1608077.38 frames. ], batch size: 20, lr: 4.64e-02, grad_scale: 16.0 +2023-02-05 18:29:54,898 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2955.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:30:21,053 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.7667, 1.7280, 5.3666, 2.6356, 5.0881, 4.7589, 4.7229, 4.6750], + device='cuda:0'), covar=tensor([0.0155, 0.3371, 0.0132, 0.1222, 0.0197, 0.0171, 0.0323, 0.0312], + device='cuda:0'), in_proj_covar=tensor([0.0078, 0.0240, 0.0099, 0.0135, 0.0115, 0.0114, 0.0121, 0.0133], + device='cuda:0'), out_proj_covar=tensor([4.9940e-05, 1.4391e-04, 6.4020e-05, 8.9970e-05, 6.7817e-05, 6.6619e-05, + 7.4881e-05, 8.3152e-05], device='cuda:0') +2023-02-05 18:30:25,906 INFO [train.py:901] (0/4) Epoch 1, batch 3000, loss[loss=0.4612, simple_loss=0.4704, pruned_loss=0.226, over 8529.00 frames. ], tot_loss[loss=0.4998, simple_loss=0.49, pruned_loss=0.2549, over 1614305.12 frames. ], batch size: 26, lr: 4.63e-02, grad_scale: 16.0 +2023-02-05 18:30:25,906 INFO [train.py:926] (0/4) Computing validation loss +2023-02-05 18:30:40,786 INFO [train.py:935] (0/4) Epoch 1, validation: loss=0.4518, simple_loss=0.5106, pruned_loss=0.1966, over 944034.00 frames. +2023-02-05 18:30:40,787 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6257MB +2023-02-05 18:30:44,887 INFO [optim.py:369] (0/4) 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 18:31:07,307 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3037.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:31:13,910 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=3047.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:31:16,514 INFO [train.py:901] (0/4) Epoch 1, batch 3050, loss[loss=0.4799, simple_loss=0.4873, pruned_loss=0.2363, over 8591.00 frames. ], tot_loss[loss=0.4965, simple_loss=0.4882, pruned_loss=0.2524, over 1617089.16 frames. ], batch size: 31, lr: 4.62e-02, grad_scale: 16.0 +2023-02-05 18:31:29,826 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.87 vs. limit=2.0 +2023-02-05 18:31:30,849 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3073.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:31:41,202 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.72 vs. limit=5.0 +2023-02-05 18:31:47,473 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3098.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:31:49,290 INFO [train.py:901] (0/4) Epoch 1, batch 3100, loss[loss=0.4634, simple_loss=0.477, pruned_loss=0.2249, over 8503.00 frames. ], tot_loss[loss=0.4965, simple_loss=0.4874, pruned_loss=0.2528, over 1617438.56 frames. ], batch size: 26, lr: 4.61e-02, grad_scale: 16.0 +2023-02-05 18:31:53,105 INFO [optim.py:369] (0/4) 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 18:32:24,771 INFO [train.py:901] (0/4) Epoch 1, batch 3150, loss[loss=0.4393, simple_loss=0.4555, pruned_loss=0.2116, over 7932.00 frames. ], tot_loss[loss=0.4936, simple_loss=0.4852, pruned_loss=0.251, over 1614670.30 frames. ], batch size: 20, lr: 4.60e-02, grad_scale: 16.0 +2023-02-05 18:32:32,211 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3162.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:32:44,602 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.7271, 4.0188, 3.4627, 1.1603, 3.3083, 3.6053, 3.4815, 2.9845], + device='cuda:0'), covar=tensor([0.1047, 0.0555, 0.0771, 0.4295, 0.0550, 0.0498, 0.1192, 0.0702], + device='cuda:0'), in_proj_covar=tensor([0.0167, 0.0123, 0.0146, 0.0208, 0.0113, 0.0094, 0.0148, 0.0109], + device='cuda:0'), out_proj_covar=tensor([1.2544e-04, 1.0251e-04, 9.8361e-05, 1.4224e-04, 7.6705e-05, 6.7290e-05, + 1.1470e-04, 7.4880e-05], device='cuda:0') +2023-02-05 18:32:47,630 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=3186.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:32:57,074 INFO [train.py:901] (0/4) Epoch 1, batch 3200, loss[loss=0.4357, simple_loss=0.4253, pruned_loss=0.223, over 7427.00 frames. ], tot_loss[loss=0.4917, simple_loss=0.4837, pruned_loss=0.2499, over 1612254.38 frames. ], batch size: 17, lr: 4.59e-02, grad_scale: 16.0 +2023-02-05 18:33:00,919 INFO [optim.py:369] (0/4) 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 18:33:32,105 INFO [train.py:901] (0/4) Epoch 1, batch 3250, loss[loss=0.567, simple_loss=0.53, pruned_loss=0.302, over 8640.00 frames. ], tot_loss[loss=0.491, simple_loss=0.483, pruned_loss=0.2495, over 1616665.62 frames. ], batch size: 49, lr: 4.58e-02, grad_scale: 16.0 +2023-02-05 18:34:04,403 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=3299.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:34:05,629 INFO [train.py:901] (0/4) Epoch 1, batch 3300, loss[loss=0.5655, simple_loss=0.5436, pruned_loss=0.2937, over 8231.00 frames. ], tot_loss[loss=0.4874, simple_loss=0.4815, pruned_loss=0.2467, over 1616771.27 frames. ], batch size: 22, lr: 4.57e-02, grad_scale: 16.0 +2023-02-05 18:34:05,826 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3301.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:34:08,942 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3306.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:34:09,426 INFO [optim.py:369] (0/4) 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 18:34:39,417 INFO [train.py:901] (0/4) Epoch 1, batch 3350, loss[loss=0.4091, simple_loss=0.4264, pruned_loss=0.1959, over 8083.00 frames. ], tot_loss[loss=0.4834, simple_loss=0.4791, pruned_loss=0.2439, over 1614974.48 frames. ], batch size: 21, lr: 4.56e-02, grad_scale: 16.0 +2023-02-05 18:34:58,220 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=2.09 vs. limit=2.0 +2023-02-05 18:35:01,938 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=3381.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:35:02,004 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.2610, 1.9015, 4.7349, 2.8178, 4.6803, 4.1182, 4.3726, 4.2091], + device='cuda:0'), covar=tensor([0.0101, 0.3281, 0.0182, 0.0875, 0.0189, 0.0218, 0.0250, 0.0281], + device='cuda:0'), in_proj_covar=tensor([0.0083, 0.0255, 0.0112, 0.0144, 0.0124, 0.0124, 0.0122, 0.0139], + device='cuda:0'), out_proj_covar=tensor([5.1228e-05, 1.5082e-04, 7.3966e-05, 9.7457e-05, 7.5781e-05, 7.5110e-05, + 7.7015e-05, 8.9081e-05], device='cuda:0') +2023-02-05 18:35:14,980 INFO [train.py:901] (0/4) Epoch 1, batch 3400, loss[loss=0.516, simple_loss=0.5062, pruned_loss=0.2629, over 8250.00 frames. ], tot_loss[loss=0.483, simple_loss=0.4794, pruned_loss=0.2433, over 1616190.23 frames. ], batch size: 24, lr: 4.55e-02, grad_scale: 16.0 +2023-02-05 18:35:19,028 INFO [optim.py:369] (0/4) 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 18:35:23,770 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3414.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:35:26,534 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3418.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:35:43,733 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3443.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:35:48,755 INFO [train.py:901] (0/4) Epoch 1, batch 3450, loss[loss=0.4862, simple_loss=0.4957, pruned_loss=0.2383, over 8367.00 frames. ], tot_loss[loss=0.4802, simple_loss=0.4775, pruned_loss=0.2414, over 1616706.84 frames. ], batch size: 24, lr: 4.54e-02, grad_scale: 16.0 +2023-02-05 18:35:54,281 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3239, 1.6848, 1.5670, 1.4659, 1.1597, 1.6066, 0.4619, 1.1602], + device='cuda:0'), covar=tensor([0.0905, 0.0694, 0.0621, 0.0668, 0.0961, 0.0786, 0.2077, 0.0964], + device='cuda:0'), in_proj_covar=tensor([0.0124, 0.0117, 0.0101, 0.0106, 0.0117, 0.0095, 0.0159, 0.0123], + device='cuda:0'), out_proj_covar=tensor([8.8045e-05, 8.9513e-05, 7.1986e-05, 7.5796e-05, 8.7248e-05, 6.6607e-05, + 1.1928e-04, 9.6599e-05], device='cuda:0') +2023-02-05 18:36:21,020 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3496.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:36:24,209 INFO [train.py:901] (0/4) Epoch 1, batch 3500, loss[loss=0.4607, simple_loss=0.4774, pruned_loss=0.222, over 8464.00 frames. ], tot_loss[loss=0.482, simple_loss=0.479, pruned_loss=0.2425, over 1615300.44 frames. ], batch size: 29, lr: 4.53e-02, grad_scale: 16.0 +2023-02-05 18:36:28,199 INFO [optim.py:369] (0/4) 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 18:36:36,243 WARNING [train.py:1067] (0/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-05 18:36:57,810 INFO [train.py:901] (0/4) Epoch 1, batch 3550, loss[loss=0.461, simple_loss=0.4805, pruned_loss=0.2208, over 8108.00 frames. ], tot_loss[loss=0.4801, simple_loss=0.478, pruned_loss=0.2411, over 1613094.82 frames. ], batch size: 23, lr: 4.51e-02, grad_scale: 16.0 +2023-02-05 18:37:02,066 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3557.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:37:07,145 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3564.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:37:19,178 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3582.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:37:33,297 INFO [train.py:901] (0/4) Epoch 1, batch 3600, loss[loss=0.4815, simple_loss=0.4953, pruned_loss=0.2338, over 8330.00 frames. ], tot_loss[loss=0.4845, simple_loss=0.48, pruned_loss=0.2445, over 1614792.57 frames. ], batch size: 25, lr: 4.50e-02, grad_scale: 16.0 +2023-02-05 18:37:37,961 INFO [optim.py:369] (0/4) 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 18:38:05,407 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0926, 1.1974, 2.1081, 0.3202, 1.7848, 1.8492, 1.1716, 2.2755], + device='cuda:0'), covar=tensor([0.0855, 0.0514, 0.0390, 0.1395, 0.0772, 0.0699, 0.1183, 0.0308], + device='cuda:0'), in_proj_covar=tensor([0.0090, 0.0074, 0.0067, 0.0091, 0.0072, 0.0077, 0.0096, 0.0071], + device='cuda:0'), out_proj_covar=tensor([6.3558e-05, 4.9461e-05, 4.6247e-05, 6.8877e-05, 5.2372e-05, 5.3235e-05, + 6.8093e-05, 4.6859e-05], device='cuda:0') +2023-02-05 18:38:06,587 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=3650.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:38:07,048 INFO [train.py:901] (0/4) Epoch 1, batch 3650, loss[loss=0.6025, simple_loss=0.5565, pruned_loss=0.3242, over 8594.00 frames. ], tot_loss[loss=0.482, simple_loss=0.4789, pruned_loss=0.2426, over 1613797.08 frames. ], batch size: 31, lr: 4.49e-02, grad_scale: 16.0 +2023-02-05 18:38:16,493 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=2.03 vs. limit=2.0 +2023-02-05 18:38:19,603 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3670.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:38:34,672 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1946, 2.1223, 1.6334, 2.6164, 1.5693, 1.3238, 1.7619, 2.0992], + device='cuda:0'), covar=tensor([0.1251, 0.1667, 0.1792, 0.0327, 0.2461, 0.2392, 0.2461, 0.1427], + device='cuda:0'), in_proj_covar=tensor([0.0242, 0.0247, 0.0236, 0.0153, 0.0301, 0.0288, 0.0322, 0.0243], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0003, 0.0002], + device='cuda:0') +2023-02-05 18:38:36,797 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3694.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:38:37,526 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3695.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:38:40,425 WARNING [train.py:1067] (0/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-05 18:38:41,126 INFO [train.py:901] (0/4) Epoch 1, batch 3700, loss[loss=0.5595, simple_loss=0.5394, pruned_loss=0.2898, over 8588.00 frames. ], tot_loss[loss=0.4825, simple_loss=0.4788, pruned_loss=0.2431, over 1604826.50 frames. ], batch size: 31, lr: 4.48e-02, grad_scale: 16.0 +2023-02-05 18:38:45,133 INFO [optim.py:369] (0/4) 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 18:39:17,444 INFO [train.py:901] (0/4) Epoch 1, batch 3750, loss[loss=0.5267, simple_loss=0.5113, pruned_loss=0.271, over 8437.00 frames. ], tot_loss[loss=0.4804, simple_loss=0.4774, pruned_loss=0.2417, over 1606863.20 frames. ], batch size: 27, lr: 4.47e-02, grad_scale: 16.0 +2023-02-05 18:39:18,333 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3752.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:39:27,111 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3765.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:39:35,221 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3777.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:39:51,672 INFO [train.py:901] (0/4) Epoch 1, batch 3800, loss[loss=0.4774, simple_loss=0.4727, pruned_loss=0.241, over 8247.00 frames. ], tot_loss[loss=0.4753, simple_loss=0.4742, pruned_loss=0.2382, over 1611789.36 frames. ], batch size: 22, lr: 4.46e-02, grad_scale: 16.0 +2023-02-05 18:39:55,874 INFO [optim.py:369] (0/4) 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 18:40:27,880 INFO [train.py:901] (0/4) Epoch 1, batch 3850, loss[loss=0.5702, simple_loss=0.5309, pruned_loss=0.3048, over 8185.00 frames. ], tot_loss[loss=0.474, simple_loss=0.4736, pruned_loss=0.2372, over 1612659.99 frames. ], batch size: 23, lr: 4.45e-02, grad_scale: 16.0 +2023-02-05 18:40:36,209 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=2.05 vs. limit=2.0 +2023-02-05 18:40:36,795 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([0.2050, 0.8129, 1.0516, 0.1129, 0.6936, 0.5957, 0.1069, 0.8981], + device='cuda:0'), covar=tensor([0.0816, 0.0510, 0.0373, 0.1089, 0.0517, 0.0761, 0.1010, 0.0518], + device='cuda:0'), in_proj_covar=tensor([0.0100, 0.0077, 0.0069, 0.0098, 0.0075, 0.0085, 0.0098, 0.0077], + device='cuda:0'), out_proj_covar=tensor([6.9988e-05, 5.1770e-05, 4.6694e-05, 7.6608e-05, 5.5707e-05, 5.9700e-05, + 7.1457e-05, 5.0921e-05], device='cuda:0') +2023-02-05 18:40:46,552 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-05 18:40:54,062 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8669, 1.6169, 3.2878, 1.5805, 2.1716, 3.7561, 3.3002, 3.0735], + device='cuda:0'), covar=tensor([0.2086, 0.2295, 0.0283, 0.2380, 0.1092, 0.0144, 0.0244, 0.0390], + device='cuda:0'), in_proj_covar=tensor([0.0216, 0.0238, 0.0133, 0.0223, 0.0172, 0.0096, 0.0098, 0.0136], + device='cuda:0'), out_proj_covar=tensor([1.6694e-04, 1.7885e-04, 1.1270e-04, 1.6122e-04, 1.4635e-04, 7.6774e-05, + 8.6735e-05, 1.0823e-04], device='cuda:0') +2023-02-05 18:40:57,951 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3390, 1.7681, 2.0464, 1.6266, 1.6451, 2.1143, 0.7323, 1.0158], + device='cuda:0'), covar=tensor([0.0963, 0.0631, 0.0450, 0.0584, 0.0641, 0.0319, 0.1783, 0.1020], + device='cuda:0'), in_proj_covar=tensor([0.0129, 0.0124, 0.0109, 0.0116, 0.0122, 0.0094, 0.0160, 0.0129], + device='cuda:0'), out_proj_covar=tensor([9.3903e-05, 9.7055e-05, 7.9660e-05, 8.4725e-05, 9.3800e-05, 6.5627e-05, + 1.2388e-04, 1.0316e-04], device='cuda:0') +2023-02-05 18:41:00,999 INFO [train.py:901] (0/4) Epoch 1, batch 3900, loss[loss=0.4413, simple_loss=0.4576, pruned_loss=0.2125, over 8643.00 frames. ], tot_loss[loss=0.4736, simple_loss=0.4736, pruned_loss=0.2368, over 1612737.06 frames. ], batch size: 34, lr: 4.44e-02, grad_scale: 16.0 +2023-02-05 18:41:02,503 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.7080, 2.3612, 2.1325, 1.9834, 2.1350, 1.9553, 2.6409, 3.0145], + device='cuda:0'), covar=tensor([0.1830, 0.2305, 0.2203, 0.2209, 0.1790, 0.2222, 0.1825, 0.1163], + device='cuda:0'), in_proj_covar=tensor([0.0252, 0.0260, 0.0235, 0.0248, 0.0259, 0.0236, 0.0252, 0.0238], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:0') +2023-02-05 18:41:04,999 INFO [optim.py:369] (0/4) 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 18:41:05,709 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=3908.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:41:29,964 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 1, batch 3950, loss[loss=0.4402, simple_loss=0.4633, pruned_loss=0.2085, over 8333.00 frames. ], tot_loss[loss=0.4726, simple_loss=0.4733, pruned_loss=0.236, over 1616131.68 frames. ], batch size: 25, lr: 4.43e-02, grad_scale: 16.0 +2023-02-05 18:42:06,739 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4195, 1.6432, 2.2104, 0.5386, 2.0235, 1.5277, 0.7526, 1.8640], + device='cuda:0'), covar=tensor([0.0855, 0.0835, 0.0376, 0.1439, 0.0696, 0.1116, 0.1366, 0.0438], + device='cuda:0'), in_proj_covar=tensor([0.0105, 0.0082, 0.0071, 0.0105, 0.0078, 0.0087, 0.0104, 0.0076], + device='cuda:0'), out_proj_covar=tensor([7.4751e-05, 5.5733e-05, 4.9549e-05, 8.0619e-05, 5.8522e-05, 6.1885e-05, + 7.5669e-05, 5.1059e-05], device='cuda:0') +2023-02-05 18:42:09,367 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-4000.pt +2023-02-05 18:42:10,929 INFO [train.py:901] (0/4) Epoch 1, batch 4000, loss[loss=0.4369, simple_loss=0.4345, pruned_loss=0.2197, over 7648.00 frames. ], tot_loss[loss=0.467, simple_loss=0.4688, pruned_loss=0.2326, over 1611562.40 frames. ], batch size: 19, lr: 4.42e-02, grad_scale: 8.0 +2023-02-05 18:42:15,527 INFO [optim.py:369] (0/4) 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 18:42:24,588 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=4021.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:42:25,860 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4023.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:42:31,864 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.67 vs. limit=5.0 +2023-02-05 18:42:36,390 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=4038.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:42:42,722 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=4046.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:42:46,104 INFO [train.py:901] (0/4) Epoch 1, batch 4050, loss[loss=0.5192, simple_loss=0.5149, pruned_loss=0.2618, over 8340.00 frames. ], tot_loss[loss=0.469, simple_loss=0.4701, pruned_loss=0.2339, over 1606781.63 frames. ], batch size: 26, lr: 4.41e-02, grad_scale: 8.0 +2023-02-05 18:43:22,349 INFO [train.py:901] (0/4) Epoch 1, batch 4100, loss[loss=0.4278, simple_loss=0.4517, pruned_loss=0.202, over 8324.00 frames. ], tot_loss[loss=0.4675, simple_loss=0.4694, pruned_loss=0.2328, over 1612152.66 frames. ], batch size: 25, lr: 4.40e-02, grad_scale: 8.0 +2023-02-05 18:43:26,892 INFO [optim.py:369] (0/4) 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 18:43:45,111 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3257, 1.2939, 1.5679, 0.1864, 1.2148, 1.0211, 0.2659, 1.2170], + device='cuda:0'), covar=tensor([0.0813, 0.0444, 0.0390, 0.1362, 0.0709, 0.0736, 0.1246, 0.0369], + device='cuda:0'), in_proj_covar=tensor([0.0109, 0.0083, 0.0073, 0.0107, 0.0078, 0.0092, 0.0105, 0.0077], + device='cuda:0'), out_proj_covar=tensor([7.6747e-05, 5.7009e-05, 5.1024e-05, 8.2382e-05, 6.0186e-05, 6.5240e-05, + 7.6998e-05, 5.1658e-05], device='cuda:0') +2023-02-05 18:43:56,550 INFO [train.py:901] (0/4) Epoch 1, batch 4150, loss[loss=0.3667, simple_loss=0.3865, pruned_loss=0.1735, over 7202.00 frames. ], tot_loss[loss=0.4642, simple_loss=0.4666, pruned_loss=0.2309, over 1606846.75 frames. ], batch size: 16, lr: 4.39e-02, grad_scale: 8.0 +2023-02-05 18:43:58,171 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4153.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:44:02,272 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=4159.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:44:20,632 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-02-05 18:44:33,534 INFO [train.py:901] (0/4) Epoch 1, batch 4200, loss[loss=0.4916, simple_loss=0.4858, pruned_loss=0.2487, over 8433.00 frames. ], tot_loss[loss=0.4601, simple_loss=0.4648, pruned_loss=0.2277, over 1609930.40 frames. ], batch size: 27, lr: 4.38e-02, grad_scale: 8.0 +2023-02-05 18:44:38,311 INFO [optim.py:369] (0/4) 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 18:44:44,323 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-05 18:45:04,973 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 1, batch 4250, loss[loss=0.3845, simple_loss=0.4233, pruned_loss=0.1728, over 8249.00 frames. ], tot_loss[loss=0.4586, simple_loss=0.4636, pruned_loss=0.2268, over 1613154.67 frames. ], batch size: 22, lr: 4.36e-02, grad_scale: 8.0 +2023-02-05 18:45:26,611 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=4279.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:45:33,325 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=4288.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:45:42,872 INFO [train.py:901] (0/4) Epoch 1, batch 4300, loss[loss=0.4075, simple_loss=0.4108, pruned_loss=0.202, over 7937.00 frames. ], tot_loss[loss=0.4578, simple_loss=0.4628, pruned_loss=0.2264, over 1612447.37 frames. ], batch size: 20, lr: 4.35e-02, grad_scale: 8.0 +2023-02-05 18:45:45,691 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=4304.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:45:47,006 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 18:46:00,776 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2962, 1.2385, 0.8557, 1.3607, 1.0532, 0.9045, 1.2194, 1.5871], + device='cuda:0'), covar=tensor([0.0950, 0.0870, 0.1881, 0.0583, 0.1354, 0.1266, 0.1294, 0.0673], + device='cuda:0'), in_proj_covar=tensor([0.0241, 0.0175, 0.0274, 0.0183, 0.0246, 0.0208, 0.0282, 0.0221], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 18:46:18,302 INFO [train.py:901] (0/4) Epoch 1, batch 4350, loss[loss=0.4512, simple_loss=0.4806, pruned_loss=0.2109, over 8471.00 frames. ], tot_loss[loss=0.4557, simple_loss=0.4613, pruned_loss=0.225, over 1609397.11 frames. ], batch size: 25, lr: 4.34e-02, grad_scale: 8.0 +2023-02-05 18:46:37,371 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-05 18:46:52,941 INFO [train.py:901] (0/4) Epoch 1, batch 4400, loss[loss=0.3878, simple_loss=0.4153, pruned_loss=0.1802, over 8204.00 frames. ], tot_loss[loss=0.4546, simple_loss=0.4607, pruned_loss=0.2243, over 1608001.59 frames. ], batch size: 23, lr: 4.33e-02, grad_scale: 8.0 +2023-02-05 18:46:54,521 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 18:46:58,921 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=4409.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 18:47:03,214 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=5.05 vs. limit=5.0 +2023-02-05 18:47:18,593 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=4434.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 18:47:21,209 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-05 18:47:28,090 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2193, 0.8903, 2.6798, 0.1928, 2.0033, 0.8544, 0.7245, 1.4920], + device='cuda:0'), covar=tensor([0.0669, 0.0560, 0.0190, 0.1216, 0.0733, 0.0702, 0.0971, 0.0327], + device='cuda:0'), in_proj_covar=tensor([0.0121, 0.0093, 0.0075, 0.0117, 0.0086, 0.0107, 0.0113, 0.0085], + device='cuda:0'), out_proj_covar=tensor([8.6563e-05, 6.3993e-05, 5.3238e-05, 9.2655e-05, 6.7449e-05, 7.6536e-05, + 8.4121e-05, 5.9770e-05], device='cuda:0') +2023-02-05 18:47:29,976 INFO [train.py:901] (0/4) Epoch 1, batch 4450, loss[loss=0.3652, simple_loss=0.3851, pruned_loss=0.1726, over 7542.00 frames. ], tot_loss[loss=0.4533, simple_loss=0.4595, pruned_loss=0.2236, over 1605207.56 frames. ], batch size: 18, lr: 4.32e-02, grad_scale: 8.0 +2023-02-05 18:47:54,978 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.81 vs. limit=2.0 +2023-02-05 18:48:04,127 INFO [train.py:901] (0/4) Epoch 1, batch 4500, loss[loss=0.4801, simple_loss=0.4771, pruned_loss=0.2415, over 8531.00 frames. ], tot_loss[loss=0.4541, simple_loss=0.4597, pruned_loss=0.2243, over 1606196.85 frames. ], batch size: 31, lr: 4.31e-02, grad_scale: 8.0 +2023-02-05 18:48:05,591 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=4503.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:48:05,701 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8032, 1.7629, 1.2806, 1.4362, 2.1669, 1.5316, 1.6289, 2.1279], + device='cuda:0'), covar=tensor([0.2097, 0.2641, 0.2976, 0.2723, 0.1735, 0.2616, 0.1951, 0.1526], + device='cuda:0'), in_proj_covar=tensor([0.0264, 0.0277, 0.0258, 0.0264, 0.0278, 0.0248, 0.0260, 0.0256], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:0') +2023-02-05 18:48:09,055 INFO [optim.py:369] (0/4) 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 18:48:09,187 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.5859, 3.8076, 3.2516, 1.4490, 3.1287, 3.2683, 3.3297, 2.7650], + device='cuda:0'), covar=tensor([0.0860, 0.0524, 0.0828, 0.3210, 0.0469, 0.0430, 0.1093, 0.0660], + device='cuda:0'), in_proj_covar=tensor([0.0218, 0.0167, 0.0189, 0.0242, 0.0146, 0.0114, 0.0186, 0.0127], + device='cuda:0'), out_proj_covar=tensor([1.6849e-04, 1.3041e-04, 1.2596e-04, 1.6317e-04, 9.7094e-05, 8.2199e-05, + 1.4205e-04, 8.9655e-05], device='cuda:0') +2023-02-05 18:48:15,347 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-05 18:48:21,817 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6260, 2.2166, 4.3664, 1.2865, 2.6704, 2.3717, 1.3721, 2.7461], + device='cuda:0'), covar=tensor([0.1124, 0.1406, 0.0153, 0.1394, 0.1087, 0.1780, 0.1253, 0.1049], + device='cuda:0'), in_proj_covar=tensor([0.0201, 0.0210, 0.0155, 0.0219, 0.0248, 0.0282, 0.0216, 0.0244], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-05 18:48:36,302 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4101, 1.3335, 3.6236, 1.5141, 1.9236, 4.7925, 4.3185, 4.2702], + device='cuda:0'), covar=tensor([0.1516, 0.2196, 0.0251, 0.2264, 0.1061, 0.0161, 0.0301, 0.0453], + device='cuda:0'), in_proj_covar=tensor([0.0228, 0.0253, 0.0135, 0.0243, 0.0180, 0.0105, 0.0106, 0.0148], + device='cuda:0'), out_proj_covar=tensor([1.8055e-04, 1.9671e-04, 1.2068e-04, 1.8267e-04, 1.6191e-04, 8.6699e-05, + 9.6650e-05, 1.2486e-04], device='cuda:0') +2023-02-05 18:48:41,814 INFO [train.py:901] (0/4) Epoch 1, batch 4550, loss[loss=0.415, simple_loss=0.447, pruned_loss=0.1915, over 8102.00 frames. ], tot_loss[loss=0.452, simple_loss=0.4589, pruned_loss=0.2226, over 1608097.08 frames. ], batch size: 23, lr: 4.30e-02, grad_scale: 8.0 +2023-02-05 18:49:16,709 INFO [train.py:901] (0/4) Epoch 1, batch 4600, loss[loss=0.4111, simple_loss=0.4143, pruned_loss=0.204, over 7720.00 frames. ], tot_loss[loss=0.4514, simple_loss=0.4586, pruned_loss=0.2221, over 1611980.98 frames. ], batch size: 18, lr: 4.29e-02, grad_scale: 8.0 +2023-02-05 18:49:21,483 INFO [optim.py:369] (0/4) 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 18:49:28,456 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4618.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:49:51,595 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 1, batch 4650, loss[loss=0.4288, simple_loss=0.46, pruned_loss=0.1988, over 8508.00 frames. ], tot_loss[loss=0.4495, simple_loss=0.4576, pruned_loss=0.2207, over 1613569.24 frames. ], batch size: 28, lr: 4.28e-02, grad_scale: 8.0 +2023-02-05 18:49:59,117 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=4659.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:50:16,202 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=4684.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:50:27,581 INFO [train.py:901] (0/4) Epoch 1, batch 4700, loss[loss=0.4419, simple_loss=0.4607, pruned_loss=0.2116, over 8615.00 frames. ], tot_loss[loss=0.4473, simple_loss=0.4567, pruned_loss=0.219, over 1619311.53 frames. ], batch size: 39, lr: 4.27e-02, grad_scale: 8.0 +2023-02-05 18:50:28,850 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.84 vs. limit=5.0 +2023-02-05 18:50:32,369 INFO [optim.py:369] (0/4) 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 18:51:01,879 INFO [train.py:901] (0/4) Epoch 1, batch 4750, loss[loss=0.4264, simple_loss=0.4264, pruned_loss=0.2131, over 5971.00 frames. ], tot_loss[loss=0.4457, simple_loss=0.4549, pruned_loss=0.2183, over 1614325.18 frames. ], batch size: 13, lr: 4.26e-02, grad_scale: 8.0 +2023-02-05 18:51:12,256 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4765.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:51:21,690 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-05 18:51:23,828 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-05 18:51:37,812 INFO [train.py:901] (0/4) Epoch 1, batch 4800, loss[loss=0.4493, simple_loss=0.4646, pruned_loss=0.217, over 8361.00 frames. ], tot_loss[loss=0.4472, simple_loss=0.4566, pruned_loss=0.2189, over 1618432.35 frames. ], batch size: 24, lr: 4.25e-02, grad_scale: 8.0 +2023-02-05 18:51:42,624 INFO [optim.py:369] (0/4) 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 18:52:11,418 INFO [train.py:901] (0/4) Epoch 1, batch 4850, loss[loss=0.424, simple_loss=0.4456, pruned_loss=0.2012, over 7937.00 frames. ], tot_loss[loss=0.4467, simple_loss=0.4556, pruned_loss=0.2189, over 1616809.59 frames. ], batch size: 20, lr: 4.24e-02, grad_scale: 8.0 +2023-02-05 18:52:13,498 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-05 18:52:27,463 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=4874.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:52:45,941 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3854, 1.1471, 2.6509, 1.3243, 1.8694, 2.8699, 2.7431, 2.5068], + device='cuda:0'), covar=tensor([0.1878, 0.2287, 0.0341, 0.2397, 0.0918, 0.0288, 0.0289, 0.0488], + device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0248, 0.0141, 0.0243, 0.0181, 0.0108, 0.0104, 0.0156], + device='cuda:0'), out_proj_covar=tensor([1.8623e-04, 1.9680e-04, 1.2941e-04, 1.8674e-04, 1.6287e-04, 9.3641e-05, + 9.6509e-05, 1.3366e-04], device='cuda:0') +2023-02-05 18:52:47,410 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 1, batch 4900, loss[loss=0.4095, simple_loss=0.432, pruned_loss=0.1935, over 8287.00 frames. ], tot_loss[loss=0.4443, simple_loss=0.4537, pruned_loss=0.2174, over 1618347.91 frames. ], batch size: 23, lr: 4.23e-02, grad_scale: 8.0 +2023-02-05 18:52:53,381 INFO [optim.py:369] (0/4) 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 18:53:10,532 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.5884, 4.8142, 4.0046, 2.3452, 3.9756, 3.9544, 4.3404, 3.3757], + device='cuda:0'), covar=tensor([0.0573, 0.0230, 0.0583, 0.2295, 0.0364, 0.0414, 0.0703, 0.0464], + device='cuda:0'), in_proj_covar=tensor([0.0221, 0.0163, 0.0186, 0.0239, 0.0146, 0.0118, 0.0191, 0.0122], + device='cuda:0'), out_proj_covar=tensor([1.7003e-04, 1.2340e-04, 1.2410e-04, 1.6094e-04, 9.8682e-05, 8.6213e-05, + 1.4129e-04, 8.6516e-05], device='cuda:0') +2023-02-05 18:53:22,701 INFO [train.py:901] (0/4) Epoch 1, batch 4950, loss[loss=0.5399, simple_loss=0.5225, pruned_loss=0.2787, over 8190.00 frames. ], tot_loss[loss=0.4413, simple_loss=0.4521, pruned_loss=0.2153, over 1618538.24 frames. ], batch size: 23, lr: 4.21e-02, grad_scale: 8.0 +2023-02-05 18:53:34,365 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0297, 2.0216, 1.5298, 1.5385, 2.0017, 1.4922, 1.8841, 2.4754], + device='cuda:0'), covar=tensor([0.1867, 0.2430, 0.2787, 0.2512, 0.1857, 0.2510, 0.1875, 0.1274], + device='cuda:0'), in_proj_covar=tensor([0.0271, 0.0284, 0.0273, 0.0273, 0.0281, 0.0255, 0.0267, 0.0261], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:0') +2023-02-05 18:53:47,766 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=2.04 vs. limit=2.0 +2023-02-05 18:53:59,098 INFO [train.py:901] (0/4) Epoch 1, batch 5000, loss[loss=0.5241, simple_loss=0.51, pruned_loss=0.2691, over 8102.00 frames. ], tot_loss[loss=0.4392, simple_loss=0.4504, pruned_loss=0.2141, over 1614843.72 frames. ], batch size: 23, lr: 4.20e-02, grad_scale: 8.0 +2023-02-05 18:54:04,631 INFO [optim.py:369] (0/4) 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 18:54:13,624 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=5021.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:54:30,611 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=5046.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 18:54:33,880 INFO [train.py:901] (0/4) Epoch 1, batch 5050, loss[loss=0.44, simple_loss=0.4351, pruned_loss=0.2225, over 7818.00 frames. ], tot_loss[loss=0.4385, simple_loss=0.4497, pruned_loss=0.2137, over 1614259.70 frames. ], batch size: 20, lr: 4.19e-02, grad_scale: 8.0 +2023-02-05 18:54:50,671 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-05 18:55:08,927 INFO [train.py:901] (0/4) Epoch 1, batch 5100, loss[loss=0.5021, simple_loss=0.4604, pruned_loss=0.2719, over 7431.00 frames. ], tot_loss[loss=0.4388, simple_loss=0.4498, pruned_loss=0.2139, over 1611802.45 frames. ], batch size: 17, lr: 4.18e-02, grad_scale: 8.0 +2023-02-05 18:55:13,606 INFO [optim.py:369] (0/4) 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 18:55:45,845 INFO [train.py:901] (0/4) Epoch 1, batch 5150, loss[loss=0.5332, simple_loss=0.5025, pruned_loss=0.282, over 6772.00 frames. ], tot_loss[loss=0.4385, simple_loss=0.4492, pruned_loss=0.2139, over 1605162.64 frames. ], batch size: 72, lr: 4.17e-02, grad_scale: 8.0 +2023-02-05 18:55:58,062 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.92 vs. limit=2.0 +2023-02-05 18:56:19,014 INFO [train.py:901] (0/4) Epoch 1, batch 5200, loss[loss=0.4029, simple_loss=0.4351, pruned_loss=0.1853, over 8298.00 frames. ], tot_loss[loss=0.4399, simple_loss=0.4503, pruned_loss=0.2148, over 1609685.69 frames. ], batch size: 23, lr: 4.16e-02, grad_scale: 8.0 +2023-02-05 18:56:23,457 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.91 vs. limit=2.0 +2023-02-05 18:56:23,573 INFO [optim.py:369] (0/4) 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 18:56:51,652 WARNING [train.py:1067] (0/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-05 18:56:55,106 INFO [train.py:901] (0/4) Epoch 1, batch 5250, loss[loss=0.4481, simple_loss=0.4551, pruned_loss=0.2205, over 8471.00 frames. ], tot_loss[loss=0.4386, simple_loss=0.4495, pruned_loss=0.2139, over 1608978.18 frames. ], batch size: 25, lr: 4.15e-02, grad_scale: 8.0 +2023-02-05 18:57:28,853 INFO [train.py:901] (0/4) Epoch 1, batch 5300, loss[loss=0.4632, simple_loss=0.475, pruned_loss=0.2257, over 8142.00 frames. ], tot_loss[loss=0.4391, simple_loss=0.4497, pruned_loss=0.2143, over 1609366.74 frames. ], batch size: 22, lr: 4.14e-02, grad_scale: 8.0 +2023-02-05 18:57:33,642 INFO [optim.py:369] (0/4) 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 18:58:04,345 INFO [train.py:901] (0/4) Epoch 1, batch 5350, loss[loss=0.4458, simple_loss=0.4611, pruned_loss=0.2152, over 8340.00 frames. ], tot_loss[loss=0.44, simple_loss=0.4505, pruned_loss=0.2148, over 1606720.97 frames. ], batch size: 26, lr: 4.13e-02, grad_scale: 8.0 +2023-02-05 18:58:39,814 INFO [train.py:901] (0/4) Epoch 1, batch 5400, loss[loss=0.5075, simple_loss=0.5052, pruned_loss=0.2549, over 8460.00 frames. ], tot_loss[loss=0.4376, simple_loss=0.4487, pruned_loss=0.2133, over 1609503.05 frames. ], batch size: 27, lr: 4.12e-02, grad_scale: 8.0 +2023-02-05 18:58:44,289 INFO [optim.py:369] (0/4) 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 18:58:55,202 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.5575, 3.8126, 3.2925, 1.3045, 3.0900, 3.2863, 3.3840, 2.9365], + device='cuda:0'), covar=tensor([0.0891, 0.0514, 0.0698, 0.3310, 0.0479, 0.0423, 0.1051, 0.0413], + device='cuda:0'), in_proj_covar=tensor([0.0239, 0.0174, 0.0198, 0.0260, 0.0159, 0.0129, 0.0196, 0.0119], + device='cuda:0'), out_proj_covar=tensor([1.8102e-04, 1.2747e-04, 1.3337e-04, 1.7181e-04, 1.0701e-04, 9.4842e-05, + 1.4412e-04, 8.4881e-05], device='cuda:0') +2023-02-05 18:59:02,359 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 +2023-02-05 18:59:13,402 INFO [train.py:901] (0/4) Epoch 1, batch 5450, loss[loss=0.4618, simple_loss=0.4716, pruned_loss=0.226, over 8362.00 frames. ], tot_loss[loss=0.4366, simple_loss=0.4481, pruned_loss=0.2125, over 1610779.66 frames. ], batch size: 24, lr: 4.11e-02, grad_scale: 8.0 +2023-02-05 18:59:15,715 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8992, 2.6305, 1.2720, 2.3480, 2.4193, 1.9101, 1.4424, 2.8317], + device='cuda:0'), covar=tensor([0.2098, 0.0912, 0.1947, 0.1071, 0.1381, 0.1440, 0.2681, 0.1106], + device='cuda:0'), in_proj_covar=tensor([0.0283, 0.0198, 0.0318, 0.0222, 0.0282, 0.0242, 0.0306, 0.0260], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 18:59:40,660 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.6373, 2.3750, 3.1723, 3.5214, 2.4714, 1.6304, 2.6761, 2.7430], + device='cuda:0'), covar=tensor([0.1863, 0.1110, 0.0380, 0.0351, 0.0825, 0.1055, 0.0760, 0.1067], + device='cuda:0'), in_proj_covar=tensor([0.0241, 0.0150, 0.0103, 0.0120, 0.0157, 0.0170, 0.0175, 0.0180], + device='cuda:0'), out_proj_covar=tensor([1.4447e-04, 9.0765e-05, 6.1179e-05, 6.9521e-05, 8.9759e-05, 1.0079e-04, + 1.0105e-04, 1.0140e-04], device='cuda:0') +2023-02-05 18:59:41,787 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 1, batch 5500, loss[loss=0.3973, simple_loss=0.4166, pruned_loss=0.189, over 7538.00 frames. ], tot_loss[loss=0.4322, simple_loss=0.4454, pruned_loss=0.2095, over 1606800.07 frames. ], batch size: 18, lr: 4.10e-02, grad_scale: 8.0 +2023-02-05 18:59:54,515 INFO [optim.py:369] (0/4) 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 19:00:20,584 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.98 vs. limit=2.0 +2023-02-05 19:00:23,629 INFO [train.py:901] (0/4) Epoch 1, batch 5550, loss[loss=0.3576, simple_loss=0.3754, pruned_loss=0.1698, over 7702.00 frames. ], tot_loss[loss=0.4335, simple_loss=0.4463, pruned_loss=0.2103, over 1609009.21 frames. ], batch size: 18, lr: 4.09e-02, grad_scale: 8.0 +2023-02-05 19:00:38,976 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9490, 2.0214, 1.0804, 2.1668, 1.7620, 1.5244, 1.4376, 2.2409], + device='cuda:0'), covar=tensor([0.1208, 0.0878, 0.2006, 0.0686, 0.1222, 0.1342, 0.1932, 0.0776], + device='cuda:0'), in_proj_covar=tensor([0.0279, 0.0194, 0.0308, 0.0220, 0.0280, 0.0238, 0.0299, 0.0252], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 19:00:59,679 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.1385, 0.9560, 2.0319, 0.2554, 1.4314, 0.9218, 0.4554, 1.5493], + device='cuda:0'), covar=tensor([0.0604, 0.0353, 0.0274, 0.0912, 0.0441, 0.0731, 0.0932, 0.0324], + device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0100, 0.0080, 0.0133, 0.0097, 0.0146, 0.0136, 0.0106], + device='cuda:0'), out_proj_covar=tensor([9.8080e-05, 7.1079e-05, 5.9889e-05, 1.0724e-04, 7.9129e-05, 1.0947e-04, + 1.0497e-04, 7.5110e-05], device='cuda:0') +2023-02-05 19:01:00,921 INFO [train.py:901] (0/4) Epoch 1, batch 5600, loss[loss=0.5119, simple_loss=0.5017, pruned_loss=0.2611, over 8479.00 frames. ], tot_loss[loss=0.435, simple_loss=0.4486, pruned_loss=0.2107, over 1613718.90 frames. ], batch size: 49, lr: 4.08e-02, grad_scale: 8.0 +2023-02-05 19:01:05,771 INFO [optim.py:369] (0/4) 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 19:01:13,361 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([0.9036, 1.4048, 1.3805, 0.0715, 1.0511, 0.9622, 0.2635, 1.4158], + device='cuda:0'), covar=tensor([0.0672, 0.0292, 0.0340, 0.1088, 0.0647, 0.0784, 0.1071, 0.0301], + device='cuda:0'), in_proj_covar=tensor([0.0140, 0.0104, 0.0083, 0.0136, 0.0101, 0.0149, 0.0141, 0.0108], + device='cuda:0'), out_proj_covar=tensor([1.0051e-04, 7.3682e-05, 6.1927e-05, 1.0937e-04, 8.2345e-05, 1.1145e-04, + 1.0865e-04, 7.6480e-05], device='cuda:0') +2023-02-05 19:01:34,544 INFO [train.py:901] (0/4) Epoch 1, batch 5650, loss[loss=0.4862, simple_loss=0.4811, pruned_loss=0.2456, over 8453.00 frames. ], tot_loss[loss=0.4354, simple_loss=0.4486, pruned_loss=0.211, over 1613895.18 frames. ], batch size: 27, lr: 4.07e-02, grad_scale: 8.0 +2023-02-05 19:01:45,699 WARNING [train.py:1067] (0/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-05 19:01:45,821 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 1, batch 5700, loss[loss=0.4094, simple_loss=0.4269, pruned_loss=0.1959, over 8680.00 frames. ], tot_loss[loss=0.4358, simple_loss=0.4483, pruned_loss=0.2117, over 1613234.54 frames. ], batch size: 49, lr: 4.06e-02, grad_scale: 8.0 +2023-02-05 19:02:15,263 INFO [optim.py:369] (0/4) 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 19:02:44,479 INFO [train.py:901] (0/4) Epoch 1, batch 5750, loss[loss=0.4475, simple_loss=0.4546, pruned_loss=0.2202, over 8607.00 frames. ], tot_loss[loss=0.4356, simple_loss=0.4489, pruned_loss=0.2111, over 1617568.36 frames. ], batch size: 31, lr: 4.05e-02, grad_scale: 8.0 +2023-02-05 19:02:51,404 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-05 19:02:59,828 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=5773.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:03:09,173 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-02-05 19:03:18,269 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0276, 2.1333, 1.8130, 2.6818, 1.8048, 1.6720, 2.0197, 2.0499], + device='cuda:0'), covar=tensor([0.0952, 0.1269, 0.1276, 0.0274, 0.1800, 0.1659, 0.1620, 0.1247], + device='cuda:0'), in_proj_covar=tensor([0.0287, 0.0314, 0.0299, 0.0177, 0.0352, 0.0336, 0.0380, 0.0289], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-05 19:03:18,566 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.59 vs. limit=5.0 +2023-02-05 19:03:19,619 INFO [train.py:901] (0/4) Epoch 1, batch 5800, loss[loss=0.4281, simple_loss=0.4452, pruned_loss=0.2055, over 8249.00 frames. ], tot_loss[loss=0.4339, simple_loss=0.4486, pruned_loss=0.2096, over 1618185.41 frames. ], batch size: 24, lr: 4.04e-02, grad_scale: 8.0 +2023-02-05 19:03:24,539 INFO [optim.py:369] (0/4) 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 19:03:57,239 INFO [train.py:901] (0/4) Epoch 1, batch 5850, loss[loss=0.4218, simple_loss=0.4454, pruned_loss=0.1991, over 8746.00 frames. ], tot_loss[loss=0.431, simple_loss=0.4466, pruned_loss=0.2077, over 1616855.27 frames. ], batch size: 30, lr: 4.03e-02, grad_scale: 8.0 +2023-02-05 19:04:15,187 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=5876.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:04:32,485 INFO [train.py:901] (0/4) Epoch 1, batch 5900, loss[loss=0.4443, simple_loss=0.4598, pruned_loss=0.2144, over 8532.00 frames. ], tot_loss[loss=0.4304, simple_loss=0.4458, pruned_loss=0.2076, over 1613178.72 frames. ], batch size: 28, lr: 4.02e-02, grad_scale: 8.0 +2023-02-05 19:04:37,231 INFO [optim.py:369] (0/4) 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 19:05:09,341 INFO [train.py:901] (0/4) Epoch 1, batch 5950, loss[loss=0.4307, simple_loss=0.4447, pruned_loss=0.2084, over 8089.00 frames. ], tot_loss[loss=0.4308, simple_loss=0.4456, pruned_loss=0.208, over 1612027.49 frames. ], batch size: 21, lr: 4.01e-02, grad_scale: 8.0 +2023-02-05 19:05:09,579 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5813, 2.0377, 3.2875, 1.1752, 2.4486, 2.0964, 1.5914, 2.1132], + device='cuda:0'), covar=tensor([0.1193, 0.1376, 0.0287, 0.1482, 0.1150, 0.1499, 0.1113, 0.1217], + device='cuda:0'), in_proj_covar=tensor([0.0269, 0.0270, 0.0234, 0.0285, 0.0337, 0.0328, 0.0279, 0.0321], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 19:05:25,872 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3268, 1.4055, 2.1769, 2.0504, 1.7555, 1.2848, 1.3817, 1.7009], + device='cuda:0'), covar=tensor([0.1642, 0.0853, 0.0287, 0.0314, 0.0500, 0.0738, 0.0677, 0.0589], + device='cuda:0'), in_proj_covar=tensor([0.0267, 0.0170, 0.0118, 0.0139, 0.0178, 0.0182, 0.0189, 0.0209], + device='cuda:0'), out_proj_covar=tensor([1.5956e-04, 1.0466e-04, 7.1609e-05, 8.2486e-05, 1.0184e-04, 1.0822e-04, + 1.0891e-04, 1.1885e-04], device='cuda:0') +2023-02-05 19:05:42,943 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-6000.pt +2023-02-05 19:05:44,547 INFO [train.py:901] (0/4) Epoch 1, batch 6000, loss[loss=0.3618, simple_loss=0.3928, pruned_loss=0.1653, over 7648.00 frames. ], tot_loss[loss=0.4289, simple_loss=0.4444, pruned_loss=0.2068, over 1612615.76 frames. ], batch size: 19, lr: 4.00e-02, grad_scale: 16.0 +2023-02-05 19:05:44,547 INFO [train.py:926] (0/4) Computing validation loss +2023-02-05 19:06:02,002 INFO [train.py:935] (0/4) Epoch 1, validation: loss=0.3351, simple_loss=0.4011, pruned_loss=0.1346, over 944034.00 frames. +2023-02-05 19:06:02,003 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6432MB +2023-02-05 19:06:06,797 INFO [optim.py:369] (0/4) 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 19:06:06,995 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6008.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:06:09,468 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6012.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 19:06:35,740 INFO [train.py:901] (0/4) Epoch 1, batch 6050, loss[loss=0.4445, simple_loss=0.4363, pruned_loss=0.2264, over 7791.00 frames. ], tot_loss[loss=0.4323, simple_loss=0.4456, pruned_loss=0.2096, over 1611324.39 frames. ], batch size: 19, lr: 3.99e-02, grad_scale: 8.0 +2023-02-05 19:06:42,855 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6061.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:07:12,055 INFO [train.py:901] (0/4) Epoch 1, batch 6100, loss[loss=0.4321, simple_loss=0.4487, pruned_loss=0.2077, over 7244.00 frames. ], tot_loss[loss=0.4295, simple_loss=0.4438, pruned_loss=0.2076, over 1609404.35 frames. ], batch size: 16, lr: 3.98e-02, grad_scale: 8.0 +2023-02-05 19:07:17,503 INFO [optim.py:369] (0/4) 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 19:07:23,140 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6117.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:07:29,001 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-05 19:07:29,786 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6127.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 19:07:45,974 INFO [train.py:901] (0/4) Epoch 1, batch 6150, loss[loss=0.455, simple_loss=0.4494, pruned_loss=0.2303, over 7185.00 frames. ], tot_loss[loss=0.4285, simple_loss=0.4426, pruned_loss=0.2072, over 1605306.05 frames. ], batch size: 71, lr: 3.97e-02, grad_scale: 8.0 +2023-02-05 19:07:47,369 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6153.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:08:03,891 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.4694, 4.7264, 4.0296, 1.7956, 3.9306, 3.9635, 4.3250, 3.6023], + device='cuda:0'), covar=tensor([0.0800, 0.0328, 0.0615, 0.3348, 0.0365, 0.0505, 0.0745, 0.0444], + device='cuda:0'), in_proj_covar=tensor([0.0244, 0.0179, 0.0209, 0.0264, 0.0164, 0.0131, 0.0199, 0.0123], + device='cuda:0'), out_proj_covar=tensor([1.8648e-04, 1.2841e-04, 1.3574e-04, 1.7370e-04, 1.0699e-04, 9.3245e-05, + 1.4131e-04, 8.7974e-05], device='cuda:0') +2023-02-05 19:08:12,159 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6188.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:08:22,924 INFO [train.py:901] (0/4) Epoch 1, batch 6200, loss[loss=0.3855, simple_loss=0.4225, pruned_loss=0.1743, over 8462.00 frames. ], tot_loss[loss=0.4293, simple_loss=0.4429, pruned_loss=0.2078, over 1603102.77 frames. ], batch size: 29, lr: 3.96e-02, grad_scale: 8.0 +2023-02-05 19:08:28,568 INFO [optim.py:369] (0/4) 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 19:08:36,379 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6220.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:08:37,158 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6221.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:08:42,653 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6229.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:08:44,631 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6232.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:08:57,419 INFO [train.py:901] (0/4) Epoch 1, batch 6250, loss[loss=0.4364, simple_loss=0.4514, pruned_loss=0.2107, over 8587.00 frames. ], tot_loss[loss=0.4298, simple_loss=0.4443, pruned_loss=0.2076, over 1609182.34 frames. ], batch size: 39, lr: 3.95e-02, grad_scale: 8.0 +2023-02-05 19:09:20,013 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6284.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:09:22,612 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.7410, 5.9164, 4.9264, 1.6712, 4.7188, 5.0888, 5.2375, 4.7758], + device='cuda:0'), covar=tensor([0.0653, 0.0271, 0.0686, 0.4229, 0.0360, 0.0485, 0.1045, 0.0428], + device='cuda:0'), in_proj_covar=tensor([0.0250, 0.0178, 0.0207, 0.0267, 0.0163, 0.0134, 0.0194, 0.0122], + device='cuda:0'), out_proj_covar=tensor([1.9024e-04, 1.2727e-04, 1.3407e-04, 1.7438e-04, 1.0577e-04, 9.5388e-05, + 1.3715e-04, 8.8490e-05], device='cuda:0') +2023-02-05 19:09:32,692 INFO [train.py:901] (0/4) Epoch 1, batch 6300, loss[loss=0.3905, simple_loss=0.4293, pruned_loss=0.1758, over 8337.00 frames. ], tot_loss[loss=0.4264, simple_loss=0.4419, pruned_loss=0.2055, over 1608993.98 frames. ], batch size: 26, lr: 3.94e-02, grad_scale: 8.0 +2023-02-05 19:09:38,769 INFO [optim.py:369] (0/4) 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 19:09:44,350 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.85 vs. limit=2.0 +2023-02-05 19:09:56,805 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6335.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:10:07,345 INFO [train.py:901] (0/4) Epoch 1, batch 6350, loss[loss=0.3989, simple_loss=0.4283, pruned_loss=0.1848, over 8254.00 frames. ], tot_loss[loss=0.4274, simple_loss=0.4421, pruned_loss=0.2064, over 1610644.43 frames. ], batch size: 24, lr: 3.93e-02, grad_scale: 8.0 +2023-02-05 19:10:08,102 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6352.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:10:28,913 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6383.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 19:10:40,801 INFO [train.py:901] (0/4) Epoch 1, batch 6400, loss[loss=0.3611, simple_loss=0.3889, pruned_loss=0.1666, over 7695.00 frames. ], tot_loss[loss=0.4264, simple_loss=0.4412, pruned_loss=0.2058, over 1606558.23 frames. ], batch size: 18, lr: 3.92e-02, grad_scale: 8.0 +2023-02-05 19:10:43,626 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6405.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:10:45,790 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 19:10:55,638 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 +2023-02-05 19:11:03,510 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4909, 1.5541, 2.5008, 1.7151, 1.5284, 1.7994, 0.5310, 1.3901], + device='cuda:0'), covar=tensor([0.0794, 0.0502, 0.0335, 0.0443, 0.0567, 0.0559, 0.1513, 0.0840], + device='cuda:0'), in_proj_covar=tensor([0.0134, 0.0106, 0.0097, 0.0126, 0.0117, 0.0084, 0.0165, 0.0140], + device='cuda:0'), out_proj_covar=tensor([1.1179e-04, 9.6465e-05, 7.9493e-05, 1.0016e-04, 1.0407e-04, 7.0098e-05, + 1.3672e-04, 1.1981e-04], device='cuda:0') +2023-02-05 19:11:16,785 INFO [train.py:901] (0/4) Epoch 1, batch 6450, loss[loss=0.5041, simple_loss=0.4796, pruned_loss=0.2643, over 7932.00 frames. ], tot_loss[loss=0.4244, simple_loss=0.4395, pruned_loss=0.2046, over 1604209.70 frames. ], batch size: 20, lr: 3.91e-02, grad_scale: 8.0 +2023-02-05 19:11:27,803 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6467.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:11:36,481 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6480.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:11:39,740 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6485.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:11:41,897 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6488.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:11:47,710 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 1, batch 6500, loss[loss=0.4135, simple_loss=0.443, pruned_loss=0.1921, over 8081.00 frames. ], tot_loss[loss=0.4226, simple_loss=0.4387, pruned_loss=0.2033, over 1602788.12 frames. ], batch size: 21, lr: 3.90e-02, grad_scale: 8.0 +2023-02-05 19:11:55,446 INFO [optim.py:369] (0/4) 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 19:11:58,508 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6513.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:12:03,286 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6520.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:12:11,142 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6532.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:12:25,065 INFO [train.py:901] (0/4) Epoch 1, batch 6550, loss[loss=0.3743, simple_loss=0.4125, pruned_loss=0.1681, over 8296.00 frames. ], tot_loss[loss=0.4229, simple_loss=0.4397, pruned_loss=0.203, over 1602634.64 frames. ], batch size: 23, lr: 3.89e-02, grad_scale: 8.0 +2023-02-05 19:12:35,937 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6565.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:12:37,958 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-05 19:12:41,463 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6573.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:12:53,791 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6591.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:12:57,648 WARNING [train.py:1067] (0/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-05 19:13:00,396 INFO [train.py:901] (0/4) Epoch 1, batch 6600, loss[loss=0.3982, simple_loss=0.4209, pruned_loss=0.1877, over 8085.00 frames. ], tot_loss[loss=0.4207, simple_loss=0.4377, pruned_loss=0.2019, over 1603069.72 frames. ], batch size: 21, lr: 3.89e-02, grad_scale: 8.0 +2023-02-05 19:13:05,684 INFO [optim.py:369] (0/4) 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 19:13:07,912 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6612.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:13:10,624 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6616.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:13:13,646 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.80 vs. limit=2.0 +2023-02-05 19:13:18,642 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6628.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:13:31,561 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6647.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:13:31,598 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6647.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:13:34,162 INFO [train.py:901] (0/4) Epoch 1, batch 6650, loss[loss=0.4454, simple_loss=0.454, pruned_loss=0.2184, over 8253.00 frames. ], tot_loss[loss=0.4189, simple_loss=0.4367, pruned_loss=0.2005, over 1608372.91 frames. ], batch size: 22, lr: 3.88e-02, grad_scale: 8.0 +2023-02-05 19:13:43,591 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8948, 1.7504, 4.0044, 1.6373, 2.3474, 5.1310, 4.2041, 4.4923], + device='cuda:0'), covar=tensor([0.1746, 0.1977, 0.0246, 0.2241, 0.0913, 0.0157, 0.0261, 0.0351], + device='cuda:0'), in_proj_covar=tensor([0.0237, 0.0251, 0.0159, 0.0241, 0.0188, 0.0120, 0.0120, 0.0172], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-05 19:13:56,271 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6680.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:14:01,301 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6688.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:14:09,932 INFO [train.py:901] (0/4) Epoch 1, batch 6700, loss[loss=0.4743, simple_loss=0.4821, pruned_loss=0.2332, over 8449.00 frames. ], tot_loss[loss=0.4197, simple_loss=0.437, pruned_loss=0.2012, over 1610222.87 frames. ], batch size: 27, lr: 3.87e-02, grad_scale: 8.0 +2023-02-05 19:14:15,399 INFO [optim.py:369] (0/4) 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 19:14:25,020 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6723.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:14:38,636 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6743.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:14:42,137 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6748.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:14:43,751 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 +2023-02-05 19:14:44,015 INFO [train.py:901] (0/4) Epoch 1, batch 6750, loss[loss=0.3832, simple_loss=0.3894, pruned_loss=0.1885, over 7309.00 frames. ], tot_loss[loss=0.4187, simple_loss=0.4365, pruned_loss=0.2004, over 1608266.37 frames. ], batch size: 16, lr: 3.86e-02, grad_scale: 8.0 +2023-02-05 19:14:45,643 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4388, 1.7946, 3.5202, 1.0551, 2.5121, 1.9931, 1.5893, 2.1658], + device='cuda:0'), covar=tensor([0.1153, 0.1651, 0.0348, 0.1658, 0.1135, 0.1810, 0.1110, 0.1480], + device='cuda:0'), in_proj_covar=tensor([0.0282, 0.0284, 0.0273, 0.0316, 0.0365, 0.0356, 0.0300, 0.0346], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 19:14:48,810 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.6347, 3.8241, 3.1547, 1.3503, 3.1217, 3.0226, 3.3661, 2.4482], + device='cuda:0'), covar=tensor([0.1022, 0.0514, 0.0949, 0.4196, 0.0563, 0.0648, 0.1132, 0.0829], + device='cuda:0'), in_proj_covar=tensor([0.0252, 0.0184, 0.0217, 0.0278, 0.0169, 0.0134, 0.0203, 0.0127], + device='cuda:0'), out_proj_covar=tensor([1.8891e-04, 1.3284e-04, 1.4250e-04, 1.8164e-04, 1.1198e-04, 9.5753e-05, + 1.3944e-04, 9.4776e-05], device='cuda:0') +2023-02-05 19:15:00,928 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6776.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:15:14,377 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-05 19:15:19,964 INFO [train.py:901] (0/4) Epoch 1, batch 6800, loss[loss=0.4147, simple_loss=0.4435, pruned_loss=0.1929, over 8486.00 frames. ], tot_loss[loss=0.4207, simple_loss=0.4386, pruned_loss=0.2014, over 1611996.73 frames. ], batch size: 29, lr: 3.85e-02, grad_scale: 8.0 +2023-02-05 19:15:20,157 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6801.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:15:25,328 INFO [optim.py:369] (0/4) 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 19:15:35,594 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6824.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:15:39,015 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6829.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:15:54,377 INFO [train.py:901] (0/4) Epoch 1, batch 6850, loss[loss=0.5122, simple_loss=0.5001, pruned_loss=0.2622, over 7150.00 frames. ], tot_loss[loss=0.4194, simple_loss=0.4376, pruned_loss=0.2006, over 1607690.62 frames. ], batch size: 71, lr: 3.84e-02, grad_scale: 8.0 +2023-02-05 19:16:04,825 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-05 19:16:05,246 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 +2023-02-05 19:16:06,405 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6868.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:16:11,624 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8521, 2.1756, 1.2429, 2.0167, 1.7302, 1.1333, 1.4547, 2.2377], + device='cuda:0'), covar=tensor([0.1286, 0.0695, 0.1576, 0.0860, 0.1350, 0.1802, 0.1925, 0.0752], + device='cuda:0'), in_proj_covar=tensor([0.0323, 0.0226, 0.0346, 0.0269, 0.0319, 0.0289, 0.0341, 0.0281], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002], + device='cuda:0') +2023-02-05 19:16:23,433 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6893.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:16:25,435 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6896.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 19:16:29,307 INFO [train.py:901] (0/4) Epoch 1, batch 6900, loss[loss=0.4649, simple_loss=0.4763, pruned_loss=0.2268, over 8500.00 frames. ], tot_loss[loss=0.4213, simple_loss=0.4396, pruned_loss=0.2015, over 1612465.38 frames. ], batch size: 26, lr: 3.83e-02, grad_scale: 8.0 +2023-02-05 19:16:31,384 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6903.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:16:35,800 INFO [optim.py:369] (0/4) 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 19:16:48,741 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6927.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:16:49,406 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6928.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:16:54,163 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7537, 2.7013, 4.5688, 1.1692, 3.0280, 2.3035, 1.9873, 2.5218], + device='cuda:0'), covar=tensor([0.1180, 0.1301, 0.0196, 0.1756, 0.1162, 0.1670, 0.1070, 0.1419], + device='cuda:0'), in_proj_covar=tensor([0.0294, 0.0296, 0.0273, 0.0324, 0.0378, 0.0353, 0.0306, 0.0359], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 19:16:54,871 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6936.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:16:56,878 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6939.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:17:00,414 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6944.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:17:00,459 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6944.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:17:03,372 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 +2023-02-05 19:17:05,068 INFO [train.py:901] (0/4) Epoch 1, batch 6950, loss[loss=0.4751, simple_loss=0.4732, pruned_loss=0.2385, over 7522.00 frames. ], tot_loss[loss=0.4231, simple_loss=0.4407, pruned_loss=0.2028, over 1610127.32 frames. ], batch size: 71, lr: 3.82e-02, grad_scale: 8.0 +2023-02-05 19:17:11,185 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-05 19:17:11,610 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.90 vs. limit=2.0 +2023-02-05 19:17:12,128 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6961.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:17:17,891 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6969.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:17:21,911 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.5587, 0.9968, 5.2835, 2.4663, 4.7035, 4.6266, 4.8982, 4.6401], + device='cuda:0'), covar=tensor([0.0190, 0.3879, 0.0175, 0.1101, 0.0736, 0.0199, 0.0214, 0.0257], + device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0331, 0.0176, 0.0205, 0.0217, 0.0183, 0.0166, 0.0197], + device='cuda:0'), out_proj_covar=tensor([9.2478e-05, 1.8279e-04, 1.0824e-04, 1.3202e-04, 1.2577e-04, 1.0947e-04, + 1.0024e-04, 1.2553e-04], device='cuda:0') +2023-02-05 19:17:32,918 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6991.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:17:38,574 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6999.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:17:39,832 INFO [train.py:901] (0/4) Epoch 1, batch 7000, loss[loss=0.4256, simple_loss=0.4553, pruned_loss=0.198, over 8586.00 frames. ], tot_loss[loss=0.4223, simple_loss=0.4407, pruned_loss=0.2019, over 1616195.53 frames. ], batch size: 31, lr: 3.81e-02, grad_scale: 8.0 +2023-02-05 19:17:45,245 INFO [optim.py:369] (0/4) 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 19:17:57,730 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 1, batch 7050, loss[loss=0.4699, simple_loss=0.4837, pruned_loss=0.228, over 8187.00 frames. ], tot_loss[loss=0.4195, simple_loss=0.4384, pruned_loss=0.2003, over 1611155.01 frames. ], batch size: 23, lr: 3.80e-02, grad_scale: 8.0 +2023-02-05 19:18:26,085 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7066.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 19:18:32,282 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0699, 1.5155, 1.9857, 0.2340, 1.5557, 1.3517, 0.3954, 1.7320], + device='cuda:0'), covar=tensor([0.0528, 0.0249, 0.0208, 0.0836, 0.0381, 0.0583, 0.0796, 0.0246], + device='cuda:0'), in_proj_covar=tensor([0.0154, 0.0124, 0.0100, 0.0165, 0.0111, 0.0180, 0.0166, 0.0125], + device='cuda:0'), out_proj_covar=tensor([1.1376e-04, 9.1073e-05, 7.9094e-05, 1.2731e-04, 9.1348e-05, 1.4211e-04, + 1.2867e-04, 9.4121e-05], device='cuda:0') +2023-02-05 19:18:34,326 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0649, 1.6803, 3.0750, 3.0811, 2.2046, 1.5750, 1.5116, 2.0499], + device='cuda:0'), covar=tensor([0.1495, 0.1149, 0.0220, 0.0221, 0.0565, 0.0691, 0.0789, 0.0892], + device='cuda:0'), in_proj_covar=tensor([0.0329, 0.0227, 0.0150, 0.0172, 0.0226, 0.0227, 0.0238, 0.0271], + device='cuda:0'), out_proj_covar=tensor([1.9455e-04, 1.4054e-04, 9.0445e-05, 9.9180e-05, 1.2754e-04, 1.3444e-04, + 1.3773e-04, 1.5255e-04], device='cuda:0') +2023-02-05 19:18:44,483 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.70 vs. limit=2.0 +2023-02-05 19:18:50,234 INFO [train.py:901] (0/4) Epoch 1, batch 7100, loss[loss=0.3659, simple_loss=0.3956, pruned_loss=0.1681, over 7821.00 frames. ], tot_loss[loss=0.4165, simple_loss=0.4363, pruned_loss=0.1983, over 1612547.62 frames. ], batch size: 19, lr: 3.79e-02, grad_scale: 8.0 +2023-02-05 19:18:53,876 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7106.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:18:55,758 INFO [optim.py:369] (0/4) 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 19:19:10,381 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([0.9308, 0.9049, 1.8784, 0.9814, 1.0455, 1.6903, 0.2971, 0.9017], + device='cuda:0'), covar=tensor([0.0635, 0.0419, 0.0277, 0.0398, 0.0458, 0.0299, 0.1122, 0.0523], + device='cuda:0'), in_proj_covar=tensor([0.0130, 0.0107, 0.0100, 0.0125, 0.0115, 0.0075, 0.0156, 0.0129], + device='cuda:0'), out_proj_covar=tensor([1.0967e-04, 9.7546e-05, 8.4730e-05, 1.0243e-04, 1.0441e-04, 6.5917e-05, + 1.3322e-04, 1.1224e-04], device='cuda:0') +2023-02-05 19:19:18,813 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.6566, 3.8603, 3.3118, 1.5576, 3.1499, 3.1784, 3.4983, 2.8510], + device='cuda:0'), covar=tensor([0.0811, 0.0463, 0.0703, 0.3179, 0.0428, 0.0563, 0.0781, 0.0455], + device='cuda:0'), in_proj_covar=tensor([0.0258, 0.0186, 0.0218, 0.0286, 0.0172, 0.0136, 0.0205, 0.0132], + device='cuda:0'), out_proj_covar=tensor([1.9254e-04, 1.3197e-04, 1.4281e-04, 1.8513e-04, 1.1163e-04, 9.7055e-05, + 1.3945e-04, 9.4535e-05], device='cuda:0') +2023-02-05 19:19:25,906 INFO [train.py:901] (0/4) Epoch 1, batch 7150, loss[loss=0.4762, simple_loss=0.4599, pruned_loss=0.2462, over 7965.00 frames. ], tot_loss[loss=0.414, simple_loss=0.4343, pruned_loss=0.1968, over 1613381.68 frames. ], batch size: 21, lr: 3.78e-02, grad_scale: 8.0 +2023-02-05 19:19:46,604 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7181.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:19:56,061 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7195.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:19:59,480 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7200.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:19:59,950 INFO [train.py:901] (0/4) Epoch 1, batch 7200, loss[loss=0.3936, simple_loss=0.4048, pruned_loss=0.1912, over 7413.00 frames. ], tot_loss[loss=0.4164, simple_loss=0.4359, pruned_loss=0.1984, over 1613380.46 frames. ], batch size: 17, lr: 3.78e-02, grad_scale: 8.0 +2023-02-05 19:20:05,331 INFO [optim.py:369] (0/4) 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 19:20:13,032 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7220.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:20:16,297 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7225.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:20:25,377 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0395, 1.2869, 2.9631, 1.0306, 1.8174, 3.1365, 3.0391, 2.8847], + device='cuda:0'), covar=tensor([0.2116, 0.1924, 0.0411, 0.2567, 0.0991, 0.0304, 0.0340, 0.0422], + device='cuda:0'), in_proj_covar=tensor([0.0245, 0.0260, 0.0165, 0.0258, 0.0188, 0.0125, 0.0124, 0.0182], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-05 19:20:25,920 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=7240.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 19:20:33,011 INFO [train.py:901] (0/4) Epoch 1, batch 7250, loss[loss=0.4558, simple_loss=0.4802, pruned_loss=0.2157, over 8198.00 frames. ], tot_loss[loss=0.4168, simple_loss=0.4365, pruned_loss=0.1985, over 1615588.43 frames. ], batch size: 23, lr: 3.77e-02, grad_scale: 8.0 +2023-02-05 19:20:48,466 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=7271.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:21:03,980 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7293.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:21:09,002 INFO [train.py:901] (0/4) Epoch 1, batch 7300, loss[loss=0.4508, simple_loss=0.4696, pruned_loss=0.216, over 8475.00 frames. ], tot_loss[loss=0.4165, simple_loss=0.437, pruned_loss=0.198, over 1620166.23 frames. ], batch size: 29, lr: 3.76e-02, grad_scale: 8.0 +2023-02-05 19:21:14,313 INFO [optim.py:369] (0/4) 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 19:21:18,594 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7976, 2.0325, 1.9002, 2.7830, 1.2163, 1.2482, 1.7216, 1.8015], + device='cuda:0'), covar=tensor([0.1266, 0.1592, 0.1321, 0.0328, 0.2252, 0.2223, 0.2239, 0.1276], + device='cuda:0'), in_proj_covar=tensor([0.0309, 0.0323, 0.0305, 0.0201, 0.0348, 0.0339, 0.0389, 0.0293], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0002, 0.0004, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-05 19:21:42,633 INFO [train.py:901] (0/4) Epoch 1, batch 7350, loss[loss=0.3431, simple_loss=0.3691, pruned_loss=0.1586, over 5963.00 frames. ], tot_loss[loss=0.4169, simple_loss=0.4373, pruned_loss=0.1983, over 1620572.95 frames. ], batch size: 13, lr: 3.75e-02, grad_scale: 8.0 +2023-02-05 19:21:45,525 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7355.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 19:21:50,123 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7362.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:21:51,639 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=2.03 vs. limit=2.0 +2023-02-05 19:21:56,009 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-05 19:22:08,446 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7386.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:22:09,144 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7387.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:22:18,307 WARNING [train.py:1067] (0/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-05 19:22:18,974 INFO [train.py:901] (0/4) Epoch 1, batch 7400, loss[loss=0.4382, simple_loss=0.4547, pruned_loss=0.2109, over 8491.00 frames. ], tot_loss[loss=0.4165, simple_loss=0.4368, pruned_loss=0.1981, over 1621267.96 frames. ], batch size: 39, lr: 3.74e-02, grad_scale: 8.0 +2023-02-05 19:22:24,409 INFO [optim.py:369] (0/4) 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 19:22:25,146 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=7410.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 19:22:35,057 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7425.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:22:50,635 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.9335, 4.0905, 3.4389, 1.5691, 3.3844, 3.3989, 3.5883, 3.1695], + device='cuda:0'), covar=tensor([0.1128, 0.0684, 0.1034, 0.4210, 0.0493, 0.0728, 0.2004, 0.0478], + device='cuda:0'), in_proj_covar=tensor([0.0266, 0.0182, 0.0224, 0.0292, 0.0173, 0.0137, 0.0208, 0.0132], + device='cuda:0'), out_proj_covar=tensor([1.9750e-04, 1.2763e-04, 1.4680e-04, 1.8740e-04, 1.1254e-04, 9.9242e-05, + 1.4099e-04, 9.3385e-05], device='cuda:0') +2023-02-05 19:22:52,553 INFO [train.py:901] (0/4) Epoch 1, batch 7450, loss[loss=0.4049, simple_loss=0.4349, pruned_loss=0.1874, over 8040.00 frames. ], tot_loss[loss=0.4154, simple_loss=0.4361, pruned_loss=0.1973, over 1618394.38 frames. ], batch size: 22, lr: 3.73e-02, grad_scale: 8.0 +2023-02-05 19:22:56,049 WARNING [train.py:1067] (0/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-05 19:23:27,518 INFO [train.py:901] (0/4) Epoch 1, batch 7500, loss[loss=0.3906, simple_loss=0.4072, pruned_loss=0.187, over 5948.00 frames. ], tot_loss[loss=0.4129, simple_loss=0.434, pruned_loss=0.1959, over 1616524.45 frames. ], batch size: 13, lr: 3.72e-02, grad_scale: 8.0 +2023-02-05 19:23:34,187 INFO [optim.py:369] (0/4) 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 19:23:45,042 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=7525.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:23:45,145 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7525.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 19:23:48,427 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.2832, 4.4540, 3.7449, 1.6352, 3.6743, 3.3560, 4.0164, 3.1112], + device='cuda:0'), covar=tensor([0.0690, 0.0357, 0.0685, 0.3265, 0.0395, 0.0623, 0.0862, 0.0471], + device='cuda:0'), in_proj_covar=tensor([0.0268, 0.0182, 0.0220, 0.0287, 0.0177, 0.0139, 0.0206, 0.0135], + device='cuda:0'), out_proj_covar=tensor([1.9864e-04, 1.2872e-04, 1.4452e-04, 1.8522e-04, 1.1414e-04, 1.0013e-04, + 1.4026e-04, 9.5095e-05], device='cuda:0') +2023-02-05 19:23:55,048 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3587, 1.9918, 1.6215, 1.4165, 2.1009, 1.7546, 2.1164, 2.2855], + device='cuda:0'), covar=tensor([0.1409, 0.1793, 0.2524, 0.2148, 0.1232, 0.1955, 0.1257, 0.1045], + device='cuda:0'), in_proj_covar=tensor([0.0271, 0.0285, 0.0297, 0.0280, 0.0264, 0.0259, 0.0257, 0.0248], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], + device='cuda:0') +2023-02-05 19:24:02,231 INFO [train.py:901] (0/4) Epoch 1, batch 7550, loss[loss=0.396, simple_loss=0.4421, pruned_loss=0.1749, over 8201.00 frames. ], tot_loss[loss=0.4111, simple_loss=0.4328, pruned_loss=0.1947, over 1618273.81 frames. ], batch size: 23, lr: 3.72e-02, grad_scale: 8.0 +2023-02-05 19:24:09,368 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 +2023-02-05 19:24:36,299 INFO [train.py:901] (0/4) Epoch 1, batch 7600, loss[loss=0.4142, simple_loss=0.4428, pruned_loss=0.1928, over 8103.00 frames. ], tot_loss[loss=0.4121, simple_loss=0.4331, pruned_loss=0.1955, over 1619382.07 frames. ], batch size: 23, lr: 3.71e-02, grad_scale: 8.0 +2023-02-05 19:24:41,739 INFO [optim.py:369] (0/4) 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 19:24:43,964 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7611.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 19:25:03,356 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7636.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 19:25:03,859 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=7637.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:25:05,940 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7642.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:25:07,358 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7642.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:25:10,528 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.8434, 5.9329, 4.9393, 1.7881, 4.8092, 5.2988, 5.4040, 4.8630], + device='cuda:0'), covar=tensor([0.0547, 0.0447, 0.0485, 0.3080, 0.0295, 0.0223, 0.0957, 0.0239], + device='cuda:0'), in_proj_covar=tensor([0.0263, 0.0178, 0.0219, 0.0281, 0.0178, 0.0135, 0.0209, 0.0133], + device='cuda:0'), out_proj_covar=tensor([1.9373e-04, 1.2532e-04, 1.4315e-04, 1.8169e-04, 1.1455e-04, 9.7686e-05, + 1.4215e-04, 9.5243e-05], device='cuda:0') +2023-02-05 19:25:13,186 INFO [train.py:901] (0/4) Epoch 1, batch 7650, loss[loss=0.3901, simple_loss=0.4256, pruned_loss=0.1773, over 8106.00 frames. ], tot_loss[loss=0.4097, simple_loss=0.4318, pruned_loss=0.1938, over 1617717.52 frames. ], batch size: 23, lr: 3.70e-02, grad_scale: 8.0 +2023-02-05 19:25:23,881 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 1, batch 7700, loss[loss=0.3548, simple_loss=0.3769, pruned_loss=0.1664, over 7536.00 frames. ], tot_loss[loss=0.4088, simple_loss=0.431, pruned_loss=0.1933, over 1619301.74 frames. ], batch size: 18, lr: 3.69e-02, grad_scale: 8.0 +2023-02-05 19:25:51,306 INFO [optim.py:369] (0/4) 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 19:25:52,867 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9765, 2.2388, 2.5784, 1.4834, 1.4755, 2.3191, 0.5874, 1.7914], + device='cuda:0'), covar=tensor([0.0480, 0.0396, 0.0278, 0.0393, 0.0580, 0.0494, 0.1368, 0.0532], + device='cuda:0'), in_proj_covar=tensor([0.0133, 0.0113, 0.0101, 0.0143, 0.0118, 0.0085, 0.0167, 0.0135], + device='cuda:0'), out_proj_covar=tensor([1.1308e-04, 1.0516e-04, 8.6391e-05, 1.1905e-04, 1.0963e-04, 7.6359e-05, + 1.4444e-04, 1.1963e-04], device='cuda:0') +2023-02-05 19:26:07,593 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 1, batch 7750, loss[loss=0.3355, simple_loss=0.3778, pruned_loss=0.1466, over 7210.00 frames. ], tot_loss[loss=0.4095, simple_loss=0.4314, pruned_loss=0.1938, over 1613678.07 frames. ], batch size: 16, lr: 3.68e-02, grad_scale: 8.0 +2023-02-05 19:26:23,165 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7752.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:26:29,120 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7761.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:26:32,510 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0450, 2.1794, 1.7701, 2.8547, 1.3023, 1.0900, 1.7199, 2.3640], + device='cuda:0'), covar=tensor([0.1294, 0.1402, 0.1451, 0.0378, 0.2300, 0.2405, 0.2306, 0.1005], + device='cuda:0'), in_proj_covar=tensor([0.0299, 0.0323, 0.0306, 0.0198, 0.0342, 0.0330, 0.0386, 0.0290], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0002, 0.0004, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-05 19:26:34,413 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=7769.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:26:42,667 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7781.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 19:26:56,374 INFO [train.py:901] (0/4) Epoch 1, batch 7800, loss[loss=0.3816, simple_loss=0.3928, pruned_loss=0.1852, over 7650.00 frames. ], tot_loss[loss=0.4092, simple_loss=0.4312, pruned_loss=0.1936, over 1615879.31 frames. ], batch size: 19, lr: 3.67e-02, grad_scale: 8.0 +2023-02-05 19:26:59,817 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7806.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 19:27:01,642 INFO [optim.py:369] (0/4) 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 19:27:09,242 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7820.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:27:29,715 INFO [train.py:901] (0/4) Epoch 1, batch 7850, loss[loss=0.3606, simple_loss=0.3854, pruned_loss=0.1679, over 7686.00 frames. ], tot_loss[loss=0.4104, simple_loss=0.4316, pruned_loss=0.1946, over 1613779.23 frames. ], batch size: 18, lr: 3.66e-02, grad_scale: 8.0 +2023-02-05 19:27:52,020 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7884.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:27:59,877 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7896.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:28:00,498 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8920, 2.1595, 1.9881, 2.8863, 1.3694, 1.3482, 1.8499, 2.2002], + device='cuda:0'), covar=tensor([0.1156, 0.1200, 0.1206, 0.0313, 0.2147, 0.2028, 0.2009, 0.1254], + device='cuda:0'), in_proj_covar=tensor([0.0301, 0.0328, 0.0315, 0.0200, 0.0341, 0.0341, 0.0388, 0.0298], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-05 19:28:01,773 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.2377, 1.0795, 3.2088, 1.1920, 2.6632, 2.6740, 2.6845, 2.8027], + device='cuda:0'), covar=tensor([0.0303, 0.2672, 0.0316, 0.1348, 0.0870, 0.0368, 0.0333, 0.0411], + device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0343, 0.0186, 0.0211, 0.0241, 0.0205, 0.0181, 0.0204], + device='cuda:0'), out_proj_covar=tensor([9.8768e-05, 1.8720e-04, 1.1075e-04, 1.3309e-04, 1.3541e-04, 1.1884e-04, + 1.0693e-04, 1.2485e-04], device='cuda:0') +2023-02-05 19:28:03,027 INFO [train.py:901] (0/4) Epoch 1, batch 7900, loss[loss=0.4486, simple_loss=0.46, pruned_loss=0.2186, over 7204.00 frames. ], tot_loss[loss=0.4098, simple_loss=0.4316, pruned_loss=0.1941, over 1613338.94 frames. ], batch size: 71, lr: 3.66e-02, grad_scale: 8.0 +2023-02-05 19:28:08,440 INFO [optim.py:369] (0/4) 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 19:28:16,371 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7921.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:28:28,389 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 +2023-02-05 19:28:35,814 INFO [train.py:901] (0/4) Epoch 1, batch 7950, loss[loss=0.4244, simple_loss=0.4465, pruned_loss=0.2012, over 8525.00 frames. ], tot_loss[loss=0.4094, simple_loss=0.4319, pruned_loss=0.1934, over 1616138.65 frames. ], batch size: 28, lr: 3.65e-02, grad_scale: 8.0 +2023-02-05 19:28:53,533 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 +2023-02-05 19:28:59,117 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=7986.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:29:07,277 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9336, 1.9678, 3.4120, 1.3462, 2.6411, 2.3853, 1.8825, 2.5455], + device='cuda:0'), covar=tensor([0.0887, 0.1256, 0.0261, 0.1521, 0.0896, 0.1096, 0.0804, 0.1132], + device='cuda:0'), in_proj_covar=tensor([0.0323, 0.0320, 0.0314, 0.0345, 0.0402, 0.0371, 0.0321, 0.0382], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 19:29:08,526 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-8000.pt +2023-02-05 19:29:10,060 INFO [train.py:901] (0/4) Epoch 1, batch 8000, loss[loss=0.4953, simple_loss=0.492, pruned_loss=0.2493, over 8279.00 frames. ], tot_loss[loss=0.4093, simple_loss=0.432, pruned_loss=0.1933, over 1616236.33 frames. ], batch size: 23, lr: 3.64e-02, grad_scale: 8.0 +2023-02-05 19:29:15,109 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=8008.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:29:15,550 INFO [optim.py:369] (0/4) 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 19:29:20,569 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=2.03 vs. limit=2.0 +2023-02-05 19:29:31,421 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8033.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:29:35,305 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.3756, 2.0224, 1.8063, 1.4791, 2.8661, 2.0530, 3.0467, 3.1157], + device='cuda:0'), covar=tensor([0.1237, 0.2588, 0.2834, 0.2503, 0.1107, 0.2344, 0.1162, 0.1068], + device='cuda:0'), in_proj_covar=tensor([0.0263, 0.0287, 0.0302, 0.0282, 0.0260, 0.0262, 0.0259, 0.0248], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], + device='cuda:0') +2023-02-05 19:29:42,078 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 +2023-02-05 19:29:43,013 INFO [train.py:901] (0/4) Epoch 1, batch 8050, loss[loss=0.3777, simple_loss=0.3897, pruned_loss=0.1829, over 7250.00 frames. ], tot_loss[loss=0.4075, simple_loss=0.43, pruned_loss=0.1925, over 1609721.87 frames. ], batch size: 16, lr: 3.63e-02, grad_scale: 16.0 +2023-02-05 19:30:05,327 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/epoch-1.pt +2023-02-05 19:30:17,284 WARNING [train.py:1067] (0/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-05 19:30:20,858 INFO [train.py:901] (0/4) Epoch 2, batch 0, loss[loss=0.4617, simple_loss=0.4599, pruned_loss=0.2317, over 8598.00 frames. ], tot_loss[loss=0.4617, simple_loss=0.4599, pruned_loss=0.2317, over 8598.00 frames. ], batch size: 31, lr: 3.56e-02, grad_scale: 8.0 +2023-02-05 19:30:20,859 INFO [train.py:926] (0/4) Computing validation loss +2023-02-05 19:30:32,400 INFO [train.py:935] (0/4) Epoch 2, validation: loss=0.3107, simple_loss=0.3861, pruned_loss=0.1176, over 944034.00 frames. +2023-02-05 19:30:32,401 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6469MB +2023-02-05 19:30:41,565 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=2.11 vs. limit=2.0 +2023-02-05 19:30:44,064 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8101.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:30:46,609 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-05 19:30:46,668 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=8105.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:30:49,934 INFO [optim.py:369] (0/4) 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 19:31:04,191 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0978, 1.4299, 1.3720, 0.3850, 1.2847, 1.1131, 0.2176, 1.4017], + device='cuda:0'), covar=tensor([0.0244, 0.0134, 0.0236, 0.0477, 0.0223, 0.0548, 0.0615, 0.0154], + device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0123, 0.0102, 0.0169, 0.0118, 0.0195, 0.0172, 0.0131], + device='cuda:0'), out_proj_covar=tensor([1.1774e-04, 8.8088e-05, 8.2302e-05, 1.2591e-04, 9.3992e-05, 1.5349e-04, + 1.3041e-04, 9.5800e-05], device='cuda:0') +2023-02-05 19:31:06,756 INFO [train.py:901] (0/4) Epoch 2, batch 50, loss[loss=0.455, simple_loss=0.4591, pruned_loss=0.2254, over 8237.00 frames. ], tot_loss[loss=0.4048, simple_loss=0.4289, pruned_loss=0.1903, over 362283.81 frames. ], batch size: 22, lr: 3.55e-02, grad_scale: 8.0 +2023-02-05 19:31:11,118 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=8140.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:31:12,615 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.91 vs. limit=5.0 +2023-02-05 19:31:20,789 WARNING [train.py:1067] (0/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-05 19:31:23,162 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=5.26 vs. limit=5.0 +2023-02-05 19:31:28,325 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=8164.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:31:29,161 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8165.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:31:41,604 INFO [train.py:901] (0/4) Epoch 2, batch 100, loss[loss=0.3335, simple_loss=0.3771, pruned_loss=0.145, over 8191.00 frames. ], tot_loss[loss=0.4035, simple_loss=0.4278, pruned_loss=0.1897, over 639893.19 frames. ], batch size: 23, lr: 3.54e-02, grad_scale: 8.0 +2023-02-05 19:31:44,284 WARNING [train.py:1067] (0/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-05 19:31:59,433 INFO [optim.py:369] (0/4) 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 19:32:06,341 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8220.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:32:15,466 INFO [train.py:901] (0/4) Epoch 2, batch 150, loss[loss=0.4276, simple_loss=0.4586, pruned_loss=0.1983, over 8341.00 frames. ], tot_loss[loss=0.4066, simple_loss=0.431, pruned_loss=0.1911, over 860906.30 frames. ], batch size: 26, lr: 3.53e-02, grad_scale: 8.0 +2023-02-05 19:32:47,318 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 +2023-02-05 19:32:47,798 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8279.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:32:50,390 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=8283.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:32:50,933 INFO [train.py:901] (0/4) Epoch 2, batch 200, loss[loss=0.376, simple_loss=0.4003, pruned_loss=0.1758, over 7811.00 frames. ], tot_loss[loss=0.4036, simple_loss=0.4282, pruned_loss=0.1895, over 1026502.68 frames. ], batch size: 20, lr: 3.52e-02, grad_scale: 8.0 +2023-02-05 19:33:08,598 INFO [optim.py:369] (0/4) 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 19:33:24,841 INFO [train.py:901] (0/4) Epoch 2, batch 250, loss[loss=0.3636, simple_loss=0.394, pruned_loss=0.1666, over 7934.00 frames. ], tot_loss[loss=0.4063, simple_loss=0.4299, pruned_loss=0.1914, over 1155217.99 frames. ], batch size: 20, lr: 3.52e-02, grad_scale: 8.0 +2023-02-05 19:33:36,305 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-05 19:33:40,664 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=8357.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:33:46,005 WARNING [train.py:1067] (0/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-05 19:33:58,438 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8382.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:33:59,606 INFO [train.py:901] (0/4) Epoch 2, batch 300, loss[loss=0.3976, simple_loss=0.4396, pruned_loss=0.1778, over 8596.00 frames. ], tot_loss[loss=0.4023, simple_loss=0.4271, pruned_loss=0.1888, over 1258728.42 frames. ], batch size: 34, lr: 3.51e-02, grad_scale: 8.0 +2023-02-05 19:34:18,661 INFO [optim.py:369] (0/4) 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 19:34:35,499 INFO [train.py:901] (0/4) Epoch 2, batch 350, loss[loss=0.407, simple_loss=0.4234, pruned_loss=0.1953, over 8609.00 frames. ], tot_loss[loss=0.4034, simple_loss=0.4279, pruned_loss=0.1894, over 1340989.84 frames. ], batch size: 34, lr: 3.50e-02, grad_scale: 8.0 +2023-02-05 19:34:43,511 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6134, 3.3278, 1.8558, 2.1950, 2.1146, 1.9995, 1.7630, 3.0018], + device='cuda:0'), covar=tensor([0.1974, 0.0503, 0.1078, 0.1339, 0.1399, 0.1211, 0.1982, 0.1202], + device='cuda:0'), in_proj_covar=tensor([0.0346, 0.0224, 0.0344, 0.0295, 0.0336, 0.0300, 0.0346, 0.0295], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:0') +2023-02-05 19:35:03,547 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=8476.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:35:09,448 INFO [train.py:901] (0/4) Epoch 2, batch 400, loss[loss=0.3779, simple_loss=0.4045, pruned_loss=0.1757, over 7792.00 frames. ], tot_loss[loss=0.4036, simple_loss=0.4278, pruned_loss=0.1897, over 1401292.78 frames. ], batch size: 19, lr: 3.49e-02, grad_scale: 8.0 +2023-02-05 19:35:20,905 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8501.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:35:27,438 INFO [optim.py:369] (0/4) 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 19:35:43,489 INFO [train.py:901] (0/4) Epoch 2, batch 450, loss[loss=0.3683, simple_loss=0.4114, pruned_loss=0.1626, over 8292.00 frames. ], tot_loss[loss=0.4026, simple_loss=0.4271, pruned_loss=0.1891, over 1449852.68 frames. ], batch size: 23, lr: 3.49e-02, grad_scale: 8.0 +2023-02-05 19:35:44,349 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=8535.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:36:01,871 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8560.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:36:17,995 INFO [train.py:901] (0/4) Epoch 2, batch 500, loss[loss=0.3658, simple_loss=0.4024, pruned_loss=0.1645, over 8237.00 frames. ], tot_loss[loss=0.401, simple_loss=0.4264, pruned_loss=0.1878, over 1487848.81 frames. ], batch size: 22, lr: 3.48e-02, grad_scale: 8.0 +2023-02-05 19:36:36,148 INFO [optim.py:369] (0/4) 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 19:36:47,548 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=8627.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:36:52,666 INFO [train.py:901] (0/4) Epoch 2, batch 550, loss[loss=0.3543, simple_loss=0.3886, pruned_loss=0.16, over 7796.00 frames. ], tot_loss[loss=0.3973, simple_loss=0.4236, pruned_loss=0.1855, over 1516448.79 frames. ], batch size: 20, lr: 3.47e-02, grad_scale: 8.0 +2023-02-05 19:37:26,521 INFO [train.py:901] (0/4) Epoch 2, batch 600, loss[loss=0.4646, simple_loss=0.4721, pruned_loss=0.2286, over 8610.00 frames. ], tot_loss[loss=0.3976, simple_loss=0.4241, pruned_loss=0.1856, over 1537952.71 frames. ], batch size: 39, lr: 3.46e-02, grad_scale: 8.0 +2023-02-05 19:37:43,312 INFO [optim.py:369] (0/4) 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 19:37:44,745 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-05 19:37:52,109 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4957, 2.2759, 1.2520, 1.9372, 2.0923, 1.4050, 1.2713, 2.2036], + device='cuda:0'), covar=tensor([0.1390, 0.0671, 0.1392, 0.0833, 0.0884, 0.1280, 0.1452, 0.0765], + device='cuda:0'), in_proj_covar=tensor([0.0349, 0.0239, 0.0352, 0.0298, 0.0337, 0.0314, 0.0346, 0.0304], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:0') +2023-02-05 19:37:59,728 INFO [train.py:901] (0/4) Epoch 2, batch 650, loss[loss=0.4464, simple_loss=0.4488, pruned_loss=0.222, over 8511.00 frames. ], tot_loss[loss=0.3958, simple_loss=0.422, pruned_loss=0.1848, over 1553681.48 frames. ], batch size: 26, lr: 3.46e-02, grad_scale: 8.0 +2023-02-05 19:38:05,396 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8742.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:38:31,188 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=8778.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:38:35,550 INFO [train.py:901] (0/4) Epoch 2, batch 700, loss[loss=0.4667, simple_loss=0.4625, pruned_loss=0.2354, over 7968.00 frames. ], tot_loss[loss=0.3954, simple_loss=0.4218, pruned_loss=0.1845, over 1567560.93 frames. ], batch size: 21, lr: 3.45e-02, grad_scale: 8.0 +2023-02-05 19:38:53,121 INFO [optim.py:369] (0/4) 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 19:38:53,977 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6504, 1.3122, 2.9010, 1.0436, 1.9629, 3.3357, 2.9457, 2.8201], + device='cuda:0'), covar=tensor([0.1369, 0.1890, 0.0431, 0.2504, 0.0839, 0.0256, 0.0364, 0.0475], + device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0253, 0.0171, 0.0249, 0.0186, 0.0137, 0.0133, 0.0193], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-05 19:39:09,181 INFO [train.py:901] (0/4) Epoch 2, batch 750, loss[loss=0.3341, simple_loss=0.3748, pruned_loss=0.1466, over 7979.00 frames. ], tot_loss[loss=0.3961, simple_loss=0.4225, pruned_loss=0.1848, over 1582762.58 frames. ], batch size: 21, lr: 3.44e-02, grad_scale: 8.0 +2023-02-05 19:39:22,679 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 +2023-02-05 19:39:26,420 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-05 19:39:35,552 WARNING [train.py:1067] (0/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-05 19:39:44,334 INFO [train.py:901] (0/4) Epoch 2, batch 800, loss[loss=0.4221, simple_loss=0.4507, pruned_loss=0.1967, over 8447.00 frames. ], tot_loss[loss=0.3989, simple_loss=0.4247, pruned_loss=0.1866, over 1594996.16 frames. ], batch size: 24, lr: 3.43e-02, grad_scale: 8.0 +2023-02-05 19:40:02,289 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 2, batch 850, loss[loss=0.3637, simple_loss=0.3781, pruned_loss=0.1746, over 7713.00 frames. ], tot_loss[loss=0.3986, simple_loss=0.4245, pruned_loss=0.1864, over 1600417.35 frames. ], batch size: 18, lr: 3.43e-02, grad_scale: 8.0 +2023-02-05 19:40:26,053 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=8945.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:40:49,494 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0701, 1.7942, 1.8302, 0.2684, 1.6793, 1.0991, 0.2411, 1.6166], + device='cuda:0'), covar=tensor([0.0343, 0.0131, 0.0170, 0.0662, 0.0250, 0.0530, 0.0614, 0.0191], + device='cuda:0'), in_proj_covar=tensor([0.0166, 0.0126, 0.0106, 0.0171, 0.0121, 0.0204, 0.0176, 0.0145], + device='cuda:0'), out_proj_covar=tensor([1.1747e-04, 8.9364e-05, 8.1633e-05, 1.2431e-04, 9.4079e-05, 1.5639e-04, + 1.3026e-04, 1.0699e-04], device='cuda:0') +2023-02-05 19:40:52,660 INFO [train.py:901] (0/4) Epoch 2, batch 900, loss[loss=0.4604, simple_loss=0.4814, pruned_loss=0.2197, over 8507.00 frames. ], tot_loss[loss=0.3956, simple_loss=0.423, pruned_loss=0.1841, over 1607216.89 frames. ], batch size: 26, lr: 3.42e-02, grad_scale: 8.0 +2023-02-05 19:41:03,966 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=8998.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:41:08,143 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9004.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:41:12,013 INFO [optim.py:369] (0/4) 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 19:41:15,319 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.23 vs. limit=5.0 +2023-02-05 19:41:21,870 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9023.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:41:27,241 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9031.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:41:29,119 INFO [train.py:901] (0/4) Epoch 2, batch 950, loss[loss=0.3412, simple_loss=0.3939, pruned_loss=0.1442, over 8356.00 frames. ], tot_loss[loss=0.3959, simple_loss=0.4233, pruned_loss=0.1843, over 1613082.77 frames. ], batch size: 24, lr: 3.41e-02, grad_scale: 8.0 +2023-02-05 19:41:57,094 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-05 19:41:58,730 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9408, 2.2247, 3.5548, 1.2256, 2.7238, 2.2597, 1.9785, 2.3487], + device='cuda:0'), covar=tensor([0.0807, 0.1062, 0.0301, 0.1475, 0.0885, 0.1348, 0.0714, 0.1240], + device='cuda:0'), in_proj_covar=tensor([0.0330, 0.0325, 0.0334, 0.0369, 0.0427, 0.0396, 0.0339, 0.0411], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:0') +2023-02-05 19:42:04,019 INFO [train.py:901] (0/4) Epoch 2, batch 1000, loss[loss=0.3647, simple_loss=0.4108, pruned_loss=0.1593, over 8197.00 frames. ], tot_loss[loss=0.394, simple_loss=0.4222, pruned_loss=0.1829, over 1613542.44 frames. ], batch size: 23, lr: 3.40e-02, grad_scale: 8.0 +2023-02-05 19:42:22,620 INFO [optim.py:369] (0/4) 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 19:42:30,648 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=9122.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:42:31,268 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-05 19:42:39,157 INFO [train.py:901] (0/4) Epoch 2, batch 1050, loss[loss=0.4084, simple_loss=0.429, pruned_loss=0.1939, over 8034.00 frames. ], tot_loss[loss=0.3939, simple_loss=0.4221, pruned_loss=0.1828, over 1614320.46 frames. ], batch size: 22, lr: 3.40e-02, grad_scale: 8.0 +2023-02-05 19:42:43,232 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-05 19:43:12,153 INFO [train.py:901] (0/4) Epoch 2, batch 1100, loss[loss=0.4099, simple_loss=0.4377, pruned_loss=0.1911, over 8494.00 frames. ], tot_loss[loss=0.3943, simple_loss=0.4219, pruned_loss=0.1834, over 1614089.15 frames. ], batch size: 28, lr: 3.39e-02, grad_scale: 8.0 +2023-02-05 19:43:30,062 INFO [optim.py:369] (0/4) 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 19:43:39,797 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 +2023-02-05 19:43:47,492 INFO [train.py:901] (0/4) Epoch 2, batch 1150, loss[loss=0.3914, simple_loss=0.4024, pruned_loss=0.1902, over 7543.00 frames. ], tot_loss[loss=0.3931, simple_loss=0.4206, pruned_loss=0.1828, over 1612810.51 frames. ], batch size: 18, lr: 3.38e-02, grad_scale: 8.0 +2023-02-05 19:43:49,769 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9237.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:43:51,007 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-05 19:44:22,137 INFO [train.py:901] (0/4) Epoch 2, batch 1200, loss[loss=0.3766, simple_loss=0.4046, pruned_loss=0.1743, over 7805.00 frames. ], tot_loss[loss=0.3919, simple_loss=0.4203, pruned_loss=0.1818, over 1612087.57 frames. ], batch size: 20, lr: 3.38e-02, grad_scale: 8.0 +2023-02-05 19:44:25,529 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=9289.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:44:41,025 INFO [optim.py:369] (0/4) 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 19:44:56,723 INFO [train.py:901] (0/4) Epoch 2, batch 1250, loss[loss=0.4183, simple_loss=0.4435, pruned_loss=0.1966, over 8143.00 frames. ], tot_loss[loss=0.3937, simple_loss=0.4212, pruned_loss=0.1831, over 1612584.14 frames. ], batch size: 22, lr: 3.37e-02, grad_scale: 4.0 +2023-02-05 19:45:07,486 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=9348.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:45:25,822 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=9375.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:45:31,822 INFO [train.py:901] (0/4) Epoch 2, batch 1300, loss[loss=0.3319, simple_loss=0.3835, pruned_loss=0.1402, over 8122.00 frames. ], tot_loss[loss=0.3904, simple_loss=0.4187, pruned_loss=0.181, over 1609873.95 frames. ], batch size: 22, lr: 3.36e-02, grad_scale: 4.0 +2023-02-05 19:45:45,732 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9404.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:45:50,302 INFO [optim.py:369] (0/4) 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 19:46:05,022 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 2, batch 1350, loss[loss=0.3632, simple_loss=0.3869, pruned_loss=0.1697, over 7290.00 frames. ], tot_loss[loss=0.388, simple_loss=0.4172, pruned_loss=0.1794, over 1613935.16 frames. ], batch size: 16, lr: 3.36e-02, grad_scale: 4.0 +2023-02-05 19:46:27,283 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9463.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:46:41,356 INFO [train.py:901] (0/4) Epoch 2, batch 1400, loss[loss=0.3931, simple_loss=0.4285, pruned_loss=0.1789, over 8607.00 frames. ], tot_loss[loss=0.3884, simple_loss=0.418, pruned_loss=0.1794, over 1616846.02 frames. ], batch size: 39, lr: 3.35e-02, grad_scale: 4.0 +2023-02-05 19:46:45,507 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9490.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:46:47,577 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9493.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:46:59,478 INFO [optim.py:369] (0/4) 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 19:47:04,243 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9518.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:47:15,060 INFO [train.py:901] (0/4) Epoch 2, batch 1450, loss[loss=0.3909, simple_loss=0.417, pruned_loss=0.1824, over 7975.00 frames. ], tot_loss[loss=0.3894, simple_loss=0.4186, pruned_loss=0.1801, over 1615037.99 frames. ], batch size: 21, lr: 3.34e-02, grad_scale: 4.0 +2023-02-05 19:47:19,054 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-05 19:47:27,883 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 +2023-02-05 19:47:39,011 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6152, 1.8507, 1.7021, 2.4627, 1.2572, 1.2036, 1.7140, 1.8874], + device='cuda:0'), covar=tensor([0.1265, 0.1156, 0.1391, 0.0489, 0.1700, 0.2197, 0.1361, 0.0978], + device='cuda:0'), in_proj_covar=tensor([0.0313, 0.0338, 0.0324, 0.0210, 0.0340, 0.0352, 0.0394, 0.0308], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0005, 0.0004], + device='cuda:0') +2023-02-05 19:47:49,250 INFO [train.py:901] (0/4) Epoch 2, batch 1500, loss[loss=0.4647, simple_loss=0.4807, pruned_loss=0.2243, over 8460.00 frames. ], tot_loss[loss=0.3909, simple_loss=0.4205, pruned_loss=0.1807, over 1617453.94 frames. ], batch size: 27, lr: 3.33e-02, grad_scale: 4.0 +2023-02-05 19:47:59,299 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8075, 2.3278, 1.5052, 2.3467, 2.0978, 1.5722, 1.7994, 2.3122], + device='cuda:0'), covar=tensor([0.1493, 0.0693, 0.1188, 0.0716, 0.0876, 0.1097, 0.1523, 0.0749], + device='cuda:0'), in_proj_covar=tensor([0.0371, 0.0248, 0.0361, 0.0302, 0.0346, 0.0316, 0.0372, 0.0308], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:0') +2023-02-05 19:48:01,348 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9602.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:48:07,896 INFO [optim.py:369] (0/4) 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 19:48:23,388 INFO [train.py:901] (0/4) Epoch 2, batch 1550, loss[loss=0.3957, simple_loss=0.4356, pruned_loss=0.1779, over 8314.00 frames. ], tot_loss[loss=0.3891, simple_loss=0.4187, pruned_loss=0.1798, over 1617764.47 frames. ], batch size: 25, lr: 3.33e-02, grad_scale: 4.0 +2023-02-05 19:48:41,689 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9660.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:48:52,351 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9676.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:48:57,577 INFO [train.py:901] (0/4) Epoch 2, batch 1600, loss[loss=0.3433, simple_loss=0.3757, pruned_loss=0.1554, over 7829.00 frames. ], tot_loss[loss=0.3889, simple_loss=0.4184, pruned_loss=0.1797, over 1617059.82 frames. ], batch size: 20, lr: 3.32e-02, grad_scale: 8.0 +2023-02-05 19:48:58,385 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9685.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:49:03,003 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.3256, 4.5789, 3.8996, 1.9318, 3.8049, 3.8520, 4.1465, 3.6138], + device='cuda:0'), covar=tensor([0.0945, 0.0402, 0.0922, 0.3471, 0.0446, 0.0483, 0.0964, 0.0388], + device='cuda:0'), in_proj_covar=tensor([0.0298, 0.0199, 0.0242, 0.0318, 0.0204, 0.0157, 0.0225, 0.0149], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:0') +2023-02-05 19:49:17,084 INFO [optim.py:369] (0/4) 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 19:49:22,879 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9719.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:49:30,586 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.98 vs. limit=2.0 +2023-02-05 19:49:33,628 INFO [train.py:901] (0/4) Epoch 2, batch 1650, loss[loss=0.4427, simple_loss=0.4632, pruned_loss=0.2111, over 8104.00 frames. ], tot_loss[loss=0.3864, simple_loss=0.4173, pruned_loss=0.1778, over 1621287.47 frames. ], batch size: 23, lr: 3.31e-02, grad_scale: 8.0 +2023-02-05 19:49:40,256 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9744.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:49:41,549 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9746.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:49:44,161 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0361, 1.8534, 1.8624, 0.2962, 1.6876, 1.4862, 0.2594, 1.8626], + device='cuda:0'), covar=tensor([0.0217, 0.0069, 0.0133, 0.0393, 0.0172, 0.0296, 0.0482, 0.0095], + device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0121, 0.0108, 0.0179, 0.0123, 0.0208, 0.0179, 0.0146], + device='cuda:0'), out_proj_covar=tensor([1.1905e-04, 8.4022e-05, 8.1694e-05, 1.2763e-04, 9.3386e-05, 1.5451e-04, + 1.2926e-04, 1.0367e-04], device='cuda:0') +2023-02-05 19:49:48,284 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3724, 1.5378, 2.3465, 1.0807, 1.8421, 1.5708, 1.4050, 1.6764], + device='cuda:0'), covar=tensor([0.0949, 0.1130, 0.0315, 0.1444, 0.0739, 0.1373, 0.0914, 0.0814], + device='cuda:0'), in_proj_covar=tensor([0.0343, 0.0335, 0.0346, 0.0373, 0.0430, 0.0399, 0.0343, 0.0415], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:0') +2023-02-05 19:49:51,536 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9761.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:49:54,946 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6786, 2.5724, 4.6603, 1.0999, 2.7540, 2.1286, 2.0714, 2.5321], + device='cuda:0'), covar=tensor([0.1092, 0.1142, 0.0327, 0.1686, 0.1159, 0.1572, 0.0817, 0.1476], + device='cuda:0'), in_proj_covar=tensor([0.0346, 0.0336, 0.0348, 0.0375, 0.0431, 0.0401, 0.0345, 0.0417], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:0') +2023-02-05 19:49:58,952 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9771.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:50:02,237 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=9776.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:50:07,334 INFO [train.py:901] (0/4) Epoch 2, batch 1700, loss[loss=0.4642, simple_loss=0.4681, pruned_loss=0.2301, over 8105.00 frames. ], tot_loss[loss=0.3875, simple_loss=0.4179, pruned_loss=0.1786, over 1620812.41 frames. ], batch size: 23, lr: 3.31e-02, grad_scale: 8.0 +2023-02-05 19:50:26,240 INFO [optim.py:369] (0/4) 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 19:50:42,246 INFO [train.py:901] (0/4) Epoch 2, batch 1750, loss[loss=0.3617, simple_loss=0.4125, pruned_loss=0.1555, over 8470.00 frames. ], tot_loss[loss=0.3875, simple_loss=0.4172, pruned_loss=0.179, over 1610571.39 frames. ], batch size: 29, lr: 3.30e-02, grad_scale: 8.0 +2023-02-05 19:50:43,029 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7360, 2.4767, 1.3887, 2.1561, 2.1177, 1.2262, 1.5044, 2.4718], + device='cuda:0'), covar=tensor([0.1484, 0.0602, 0.1428, 0.0797, 0.1045, 0.1451, 0.1696, 0.0822], + device='cuda:0'), in_proj_covar=tensor([0.0373, 0.0246, 0.0365, 0.0299, 0.0351, 0.0312, 0.0378, 0.0315], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:0') +2023-02-05 19:51:16,311 INFO [train.py:901] (0/4) Epoch 2, batch 1800, loss[loss=0.3725, simple_loss=0.3824, pruned_loss=0.1813, over 7435.00 frames. ], tot_loss[loss=0.3867, simple_loss=0.4162, pruned_loss=0.1786, over 1604682.25 frames. ], batch size: 17, lr: 3.29e-02, grad_scale: 8.0 +2023-02-05 19:51:21,255 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9891.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:51:34,074 INFO [optim.py:369] (0/4) 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 19:51:49,945 INFO [train.py:901] (0/4) Epoch 2, batch 1850, loss[loss=0.4174, simple_loss=0.4369, pruned_loss=0.199, over 8136.00 frames. ], tot_loss[loss=0.3863, simple_loss=0.4156, pruned_loss=0.1785, over 1605015.13 frames. ], batch size: 22, lr: 3.29e-02, grad_scale: 8.0 +2023-02-05 19:51:58,649 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=9946.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:52:24,360 INFO [train.py:901] (0/4) Epoch 2, batch 1900, loss[loss=0.3523, simple_loss=0.3846, pruned_loss=0.16, over 8089.00 frames. ], tot_loss[loss=0.3854, simple_loss=0.4155, pruned_loss=0.1777, over 1609531.42 frames. ], batch size: 21, lr: 3.28e-02, grad_scale: 8.0 +2023-02-05 19:52:35,047 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-10000.pt +2023-02-05 19:52:43,411 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.1488, 2.0399, 1.8119, 0.3585, 1.9216, 1.2850, 0.3281, 1.9903], + device='cuda:0'), covar=tensor([0.0220, 0.0059, 0.0166, 0.0270, 0.0134, 0.0369, 0.0375, 0.0084], + device='cuda:0'), in_proj_covar=tensor([0.0177, 0.0125, 0.0111, 0.0168, 0.0125, 0.0217, 0.0181, 0.0152], + device='cuda:0'), out_proj_covar=tensor([1.2202e-04, 8.6516e-05, 8.1972e-05, 1.1786e-04, 9.3166e-05, 1.5942e-04, + 1.2926e-04, 1.0817e-04], device='cuda:0') +2023-02-05 19:52:43,808 INFO [optim.py:369] (0/4) 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 19:52:48,221 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 +2023-02-05 19:52:49,913 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=10020.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:52:50,742 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2099, 1.9521, 1.7660, 0.2105, 1.6491, 1.2285, 0.2370, 2.0189], + device='cuda:0'), covar=tensor([0.0198, 0.0065, 0.0147, 0.0283, 0.0150, 0.0418, 0.0391, 0.0075], + device='cuda:0'), in_proj_covar=tensor([0.0176, 0.0126, 0.0109, 0.0167, 0.0125, 0.0216, 0.0180, 0.0152], + device='cuda:0'), out_proj_covar=tensor([1.2150e-04, 8.6732e-05, 8.0194e-05, 1.1712e-04, 9.3715e-05, 1.5911e-04, + 1.2816e-04, 1.0794e-04], device='cuda:0') +2023-02-05 19:52:54,616 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-05 19:52:59,200 INFO [train.py:901] (0/4) Epoch 2, batch 1950, loss[loss=0.3697, simple_loss=0.4148, pruned_loss=0.1623, over 8350.00 frames. ], tot_loss[loss=0.3849, simple_loss=0.415, pruned_loss=0.1774, over 1610990.95 frames. ], batch size: 24, lr: 3.27e-02, grad_scale: 8.0 +2023-02-05 19:53:06,860 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-05 19:53:15,918 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10057.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:53:19,405 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10061.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:53:25,592 WARNING [train.py:1067] (0/4) 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] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10080.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 19:53:35,478 INFO [train.py:901] (0/4) Epoch 2, batch 2000, loss[loss=0.4506, simple_loss=0.4647, pruned_loss=0.2182, over 8470.00 frames. ], tot_loss[loss=0.3874, simple_loss=0.4175, pruned_loss=0.1787, over 1618961.14 frames. ], batch size: 25, lr: 3.27e-02, grad_scale: 8.0 +2023-02-05 19:53:46,356 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2256, 1.4479, 2.2726, 0.9468, 1.9841, 1.4743, 1.3113, 1.6697], + device='cuda:0'), covar=tensor([0.1302, 0.1354, 0.0417, 0.1941, 0.0783, 0.1741, 0.1130, 0.1010], + device='cuda:0'), in_proj_covar=tensor([0.0350, 0.0335, 0.0352, 0.0387, 0.0435, 0.0401, 0.0344, 0.0424], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:0') +2023-02-05 19:53:50,427 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=10105.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:53:55,730 INFO [optim.py:369] (0/4) 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 19:54:10,563 INFO [train.py:901] (0/4) Epoch 2, batch 2050, loss[loss=0.3765, simple_loss=0.4255, pruned_loss=0.1638, over 8352.00 frames. ], tot_loss[loss=0.3871, simple_loss=0.4166, pruned_loss=0.1788, over 1617347.19 frames. ], batch size: 24, lr: 3.26e-02, grad_scale: 4.0 +2023-02-05 19:54:11,450 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10147.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:54:36,879 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10172.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:54:45,485 INFO [train.py:901] (0/4) Epoch 2, batch 2100, loss[loss=0.3905, simple_loss=0.4436, pruned_loss=0.1687, over 8505.00 frames. ], tot_loss[loss=0.3852, simple_loss=0.4156, pruned_loss=0.1774, over 1617483.95 frames. ], batch size: 26, lr: 3.25e-02, grad_scale: 4.0 +2023-02-05 19:55:06,153 INFO [optim.py:369] (0/4) 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 19:55:11,270 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10220.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:55:20,261 INFO [train.py:901] (0/4) Epoch 2, batch 2150, loss[loss=0.3699, simple_loss=0.4072, pruned_loss=0.1663, over 8598.00 frames. ], tot_loss[loss=0.3834, simple_loss=0.4145, pruned_loss=0.1761, over 1615640.36 frames. ], batch size: 48, lr: 3.25e-02, grad_scale: 4.0 +2023-02-05 19:55:50,778 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.1019, 2.6442, 3.8800, 4.2853, 2.8756, 2.2310, 1.9659, 2.5310], + device='cuda:0'), covar=tensor([0.0713, 0.0763, 0.0145, 0.0177, 0.0398, 0.0384, 0.0568, 0.0709], + device='cuda:0'), in_proj_covar=tensor([0.0445, 0.0354, 0.0255, 0.0296, 0.0375, 0.0330, 0.0351, 0.0398], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 19:55:53,983 INFO [train.py:901] (0/4) Epoch 2, batch 2200, loss[loss=0.3833, simple_loss=0.4156, pruned_loss=0.1755, over 8046.00 frames. ], tot_loss[loss=0.3851, simple_loss=0.4154, pruned_loss=0.1774, over 1613885.37 frames. ], batch size: 22, lr: 3.24e-02, grad_scale: 4.0 +2023-02-05 19:56:06,177 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7714, 1.9815, 3.6078, 1.1542, 2.7014, 2.2367, 1.7619, 2.0035], + device='cuda:0'), covar=tensor([0.0904, 0.1215, 0.0281, 0.1697, 0.0850, 0.1280, 0.0871, 0.1359], + device='cuda:0'), in_proj_covar=tensor([0.0351, 0.0345, 0.0366, 0.0395, 0.0445, 0.0406, 0.0356, 0.0433], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:0') +2023-02-05 19:56:14,591 INFO [optim.py:369] (0/4) 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 19:56:18,156 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10317.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:56:29,263 INFO [train.py:901] (0/4) Epoch 2, batch 2250, loss[loss=0.386, simple_loss=0.4246, pruned_loss=0.1737, over 8453.00 frames. ], tot_loss[loss=0.3872, simple_loss=0.4171, pruned_loss=0.1787, over 1615145.18 frames. ], batch size: 27, lr: 3.24e-02, grad_scale: 4.0 +2023-02-05 19:56:34,609 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10342.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:57:03,211 INFO [train.py:901] (0/4) Epoch 2, batch 2300, loss[loss=0.4557, simple_loss=0.4648, pruned_loss=0.2234, over 7128.00 frames. ], tot_loss[loss=0.3858, simple_loss=0.4157, pruned_loss=0.1779, over 1605823.81 frames. ], batch size: 72, lr: 3.23e-02, grad_scale: 4.0 +2023-02-05 19:57:08,297 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10391.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:57:15,020 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=10401.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:57:23,809 INFO [optim.py:369] (0/4) 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 19:57:25,921 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10416.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:57:31,788 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=10424.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 19:57:33,881 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10427.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:57:39,139 INFO [train.py:901] (0/4) Epoch 2, batch 2350, loss[loss=0.3116, simple_loss=0.3526, pruned_loss=0.1353, over 7687.00 frames. ], tot_loss[loss=0.3845, simple_loss=0.4148, pruned_loss=0.1771, over 1606883.75 frames. ], batch size: 18, lr: 3.22e-02, grad_scale: 4.0 +2023-02-05 19:58:05,149 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10472.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:58:07,801 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10476.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:58:12,897 INFO [train.py:901] (0/4) Epoch 2, batch 2400, loss[loss=0.3516, simple_loss=0.3849, pruned_loss=0.1591, over 7719.00 frames. ], tot_loss[loss=0.3832, simple_loss=0.414, pruned_loss=0.1762, over 1607972.56 frames. ], batch size: 18, lr: 3.22e-02, grad_scale: 8.0 +2023-02-05 19:58:24,664 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 19:58:34,724 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10516.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 19:58:47,330 INFO [train.py:901] (0/4) Epoch 2, batch 2450, loss[loss=0.3721, simple_loss=0.4156, pruned_loss=0.1643, over 8336.00 frames. ], tot_loss[loss=0.3854, simple_loss=0.4157, pruned_loss=0.1776, over 1611519.61 frames. ], batch size: 26, lr: 3.21e-02, grad_scale: 8.0 +2023-02-05 19:58:50,867 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10539.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 19:58:53,810 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.29 vs. limit=5.0 +2023-02-05 19:59:22,252 INFO [train.py:901] (0/4) Epoch 2, batch 2500, loss[loss=0.3596, simple_loss=0.3976, pruned_loss=0.1608, over 8511.00 frames. ], tot_loss[loss=0.3827, simple_loss=0.4136, pruned_loss=0.1759, over 1613363.48 frames. ], batch size: 28, lr: 3.20e-02, grad_scale: 8.0 +2023-02-05 19:59:42,162 INFO [optim.py:369] (0/4) 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 19:59:43,168 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 +2023-02-05 19:59:55,951 INFO [train.py:901] (0/4) Epoch 2, batch 2550, loss[loss=0.4213, simple_loss=0.4369, pruned_loss=0.2029, over 8484.00 frames. ], tot_loss[loss=0.3835, simple_loss=0.414, pruned_loss=0.1765, over 1612684.25 frames. ], batch size: 26, lr: 3.20e-02, grad_scale: 8.0 +2023-02-05 20:00:31,355 INFO [train.py:901] (0/4) Epoch 2, batch 2600, loss[loss=0.4055, simple_loss=0.4125, pruned_loss=0.1993, over 7806.00 frames. ], tot_loss[loss=0.3838, simple_loss=0.4144, pruned_loss=0.1766, over 1610797.26 frames. ], batch size: 19, lr: 3.19e-02, grad_scale: 8.0 +2023-02-05 20:00:50,499 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 2, batch 2650, loss[loss=0.3706, simple_loss=0.4113, pruned_loss=0.1649, over 8543.00 frames. ], tot_loss[loss=0.3836, simple_loss=0.4148, pruned_loss=0.1762, over 1614975.55 frames. ], batch size: 49, lr: 3.19e-02, grad_scale: 8.0 +2023-02-05 20:01:24,204 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10762.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:01:30,872 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=10771.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:01:31,714 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10772.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:01:40,306 INFO [train.py:901] (0/4) Epoch 2, batch 2700, loss[loss=0.3779, simple_loss=0.3991, pruned_loss=0.1784, over 7798.00 frames. ], tot_loss[loss=0.3836, simple_loss=0.4145, pruned_loss=0.1763, over 1616409.37 frames. ], batch size: 19, lr: 3.18e-02, grad_scale: 8.0 +2023-02-05 20:01:46,654 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10792.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:01:48,795 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10797.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:02:01,038 INFO [optim.py:369] (0/4) 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 20:02:03,290 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=10816.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:02:06,079 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 2, batch 2750, loss[loss=0.3862, simple_loss=0.4029, pruned_loss=0.1848, over 7975.00 frames. ], tot_loss[loss=0.3825, simple_loss=0.4136, pruned_loss=0.1757, over 1613564.91 frames. ], batch size: 21, lr: 3.17e-02, grad_scale: 8.0 +2023-02-05 20:02:49,765 INFO [train.py:901] (0/4) Epoch 2, batch 2800, loss[loss=0.3605, simple_loss=0.3614, pruned_loss=0.1798, over 6798.00 frames. ], tot_loss[loss=0.3809, simple_loss=0.4124, pruned_loss=0.1747, over 1607136.97 frames. ], batch size: 15, lr: 3.17e-02, grad_scale: 8.0 +2023-02-05 20:02:51,254 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10886.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:03:03,313 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 20:03:23,111 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10931.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:03:25,018 INFO [train.py:901] (0/4) Epoch 2, batch 2850, loss[loss=0.4083, simple_loss=0.4299, pruned_loss=0.1934, over 8677.00 frames. ], tot_loss[loss=0.3795, simple_loss=0.4118, pruned_loss=0.1736, over 1612744.78 frames. ], batch size: 39, lr: 3.16e-02, grad_scale: 8.0 +2023-02-05 20:03:59,104 INFO [train.py:901] (0/4) Epoch 2, batch 2900, loss[loss=0.4482, simple_loss=0.4656, pruned_loss=0.2154, over 8416.00 frames. ], tot_loss[loss=0.3782, simple_loss=0.4112, pruned_loss=0.1726, over 1614932.58 frames. ], batch size: 29, lr: 3.16e-02, grad_scale: 8.0 +2023-02-05 20:04:19,452 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 2, batch 2950, loss[loss=0.5481, simple_loss=0.522, pruned_loss=0.2871, over 8490.00 frames. ], tot_loss[loss=0.3792, simple_loss=0.4114, pruned_loss=0.1735, over 1613129.81 frames. ], batch size: 29, lr: 3.15e-02, grad_scale: 8.0 +2023-02-05 20:04:39,270 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-05 20:04:50,967 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7488, 2.0629, 1.8769, 2.8261, 1.4984, 1.2294, 1.7354, 2.1865], + device='cuda:0'), covar=tensor([0.1364, 0.1513, 0.1615, 0.0424, 0.2073, 0.2560, 0.2243, 0.1175], + device='cuda:0'), in_proj_covar=tensor([0.0301, 0.0322, 0.0307, 0.0207, 0.0317, 0.0323, 0.0368, 0.0297], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], + device='cuda:0') +2023-02-05 20:05:08,646 INFO [train.py:901] (0/4) Epoch 2, batch 3000, loss[loss=0.4348, simple_loss=0.4418, pruned_loss=0.2139, over 6811.00 frames. ], tot_loss[loss=0.3799, simple_loss=0.4118, pruned_loss=0.174, over 1612671.16 frames. ], batch size: 72, lr: 3.14e-02, grad_scale: 8.0 +2023-02-05 20:05:08,646 INFO [train.py:926] (0/4) Computing validation loss +2023-02-05 20:05:24,857 INFO [train.py:935] (0/4) Epoch 2, validation: loss=0.2878, simple_loss=0.369, pruned_loss=0.1033, over 944034.00 frames. +2023-02-05 20:05:24,859 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6484MB +2023-02-05 20:05:27,734 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.0461, 1.1556, 3.2353, 1.0529, 2.7103, 2.7882, 2.8298, 2.8756], + device='cuda:0'), covar=tensor([0.0365, 0.2540, 0.0321, 0.1420, 0.0912, 0.0374, 0.0364, 0.0454], + device='cuda:0'), in_proj_covar=tensor([0.0178, 0.0357, 0.0220, 0.0249, 0.0296, 0.0231, 0.0219, 0.0249], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001], + device='cuda:0') +2023-02-05 20:05:39,940 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0678, 1.2233, 1.4210, 1.0779, 0.8939, 1.4108, 0.1025, 0.7151], + device='cuda:0'), covar=tensor([0.0820, 0.0461, 0.0259, 0.0529, 0.0674, 0.0298, 0.1464, 0.0668], + device='cuda:0'), in_proj_covar=tensor([0.0122, 0.0098, 0.0085, 0.0135, 0.0119, 0.0083, 0.0159, 0.0126], + device='cuda:0'), out_proj_covar=tensor([1.1463e-04, 1.0033e-04, 8.0509e-05, 1.2561e-04, 1.1919e-04, 7.8357e-05, + 1.4852e-04, 1.2342e-04], device='cuda:0') +2023-02-05 20:05:40,474 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) Epoch 2, batch 3050, loss[loss=0.3314, simple_loss=0.3756, pruned_loss=0.1436, over 7653.00 frames. ], tot_loss[loss=0.3806, simple_loss=0.4121, pruned_loss=0.1745, over 1609645.38 frames. ], batch size: 19, lr: 3.14e-02, grad_scale: 8.0 +2023-02-05 20:06:01,558 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=11136.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:06:05,881 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11142.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:06:14,855 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.5141, 2.1666, 3.3301, 3.0916, 2.8870, 2.1865, 1.5081, 2.3187], + device='cuda:0'), covar=tensor([0.0645, 0.0786, 0.0141, 0.0184, 0.0277, 0.0307, 0.0532, 0.0568], + device='cuda:0'), in_proj_covar=tensor([0.0463, 0.0382, 0.0270, 0.0310, 0.0399, 0.0343, 0.0367, 0.0411], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 20:06:23,841 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 +2023-02-05 20:06:24,312 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11167.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:06:35,407 INFO [train.py:901] (0/4) Epoch 2, batch 3100, loss[loss=0.3631, simple_loss=0.3995, pruned_loss=0.1634, over 7974.00 frames. ], tot_loss[loss=0.3789, simple_loss=0.4115, pruned_loss=0.1732, over 1613396.45 frames. ], batch size: 21, lr: 3.13e-02, grad_scale: 8.0 +2023-02-05 20:06:37,612 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11187.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:06:40,548 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 +2023-02-05 20:06:40,860 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11192.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:06:55,382 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11212.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:06:55,866 INFO [optim.py:369] (0/4) 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 20:07:01,724 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11221.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:07:10,341 INFO [train.py:901] (0/4) Epoch 2, batch 3150, loss[loss=0.3331, simple_loss=0.3754, pruned_loss=0.1454, over 8356.00 frames. ], tot_loss[loss=0.3779, simple_loss=0.4103, pruned_loss=0.1728, over 1610006.04 frames. ], batch size: 24, lr: 3.13e-02, grad_scale: 8.0 +2023-02-05 20:07:20,131 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=11247.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 20:07:22,964 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11251.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:07:46,072 INFO [train.py:901] (0/4) Epoch 2, batch 3200, loss[loss=0.3618, simple_loss=0.4088, pruned_loss=0.1574, over 8362.00 frames. ], tot_loss[loss=0.3777, simple_loss=0.4103, pruned_loss=0.1725, over 1612463.87 frames. ], batch size: 24, lr: 3.12e-02, grad_scale: 8.0 +2023-02-05 20:08:06,196 INFO [optim.py:369] (0/4) 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 20:08:21,236 INFO [train.py:901] (0/4) Epoch 2, batch 3250, loss[loss=0.3397, simple_loss=0.3831, pruned_loss=0.1482, over 8240.00 frames. ], tot_loss[loss=0.3783, simple_loss=0.4113, pruned_loss=0.1726, over 1615898.45 frames. ], batch size: 22, lr: 3.11e-02, grad_scale: 8.0 +2023-02-05 20:08:39,994 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11362.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 20:08:55,024 INFO [train.py:901] (0/4) Epoch 2, batch 3300, loss[loss=0.3985, simple_loss=0.4346, pruned_loss=0.1812, over 8478.00 frames. ], tot_loss[loss=0.38, simple_loss=0.4127, pruned_loss=0.1737, over 1617516.86 frames. ], batch size: 29, lr: 3.11e-02, grad_scale: 8.0 +2023-02-05 20:09:16,005 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 2, batch 3350, loss[loss=0.3563, simple_loss=0.3862, pruned_loss=0.1632, over 7807.00 frames. ], tot_loss[loss=0.3784, simple_loss=0.412, pruned_loss=0.1723, over 1622531.82 frames. ], batch size: 20, lr: 3.10e-02, grad_scale: 8.0 +2023-02-05 20:09:40,002 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 +2023-02-05 20:09:53,262 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4803, 2.4170, 2.7264, 1.0770, 3.0497, 1.9687, 0.9517, 1.7091], + device='cuda:0'), covar=tensor([0.0245, 0.0088, 0.0336, 0.0240, 0.0127, 0.0271, 0.0423, 0.0178], + device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0124, 0.0124, 0.0182, 0.0131, 0.0230, 0.0196, 0.0168], + device='cuda:0'), out_proj_covar=tensor([1.2343e-04, 8.0534e-05, 8.5780e-05, 1.1887e-04, 9.0968e-05, 1.6021e-04, + 1.3265e-04, 1.1235e-04], device='cuda:0') +2023-02-05 20:10:00,582 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11477.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:10:05,260 INFO [train.py:901] (0/4) Epoch 2, batch 3400, loss[loss=0.3609, simple_loss=0.3932, pruned_loss=0.1643, over 7920.00 frames. ], tot_loss[loss=0.3768, simple_loss=0.4112, pruned_loss=0.1712, over 1621409.78 frames. ], batch size: 20, lr: 3.10e-02, grad_scale: 8.0 +2023-02-05 20:10:11,476 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.8461, 1.6136, 5.7223, 2.4485, 5.1284, 4.8378, 5.2747, 5.3173], + device='cuda:0'), covar=tensor([0.0254, 0.2696, 0.0150, 0.0979, 0.0594, 0.0206, 0.0176, 0.0227], + device='cuda:0'), in_proj_covar=tensor([0.0179, 0.0363, 0.0221, 0.0250, 0.0308, 0.0243, 0.0224, 0.0255], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001], + device='cuda:0') +2023-02-05 20:10:17,702 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11502.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:10:21,203 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11507.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:10:25,777 INFO [optim.py:369] (0/4) 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 20:10:39,490 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11532.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:10:40,652 INFO [train.py:901] (0/4) Epoch 2, batch 3450, loss[loss=0.4436, simple_loss=0.4613, pruned_loss=0.213, over 8467.00 frames. ], tot_loss[loss=0.3781, simple_loss=0.4119, pruned_loss=0.1721, over 1619113.85 frames. ], batch size: 25, lr: 3.09e-02, grad_scale: 8.0 +2023-02-05 20:10:42,064 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=11536.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:10:51,630 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11550.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:11:13,179 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 +2023-02-05 20:11:15,647 INFO [train.py:901] (0/4) Epoch 2, batch 3500, loss[loss=0.346, simple_loss=0.3965, pruned_loss=0.1478, over 8509.00 frames. ], tot_loss[loss=0.376, simple_loss=0.4105, pruned_loss=0.1708, over 1615781.24 frames. ], batch size: 26, lr: 3.09e-02, grad_scale: 8.0 +2023-02-05 20:11:18,536 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2075, 1.8277, 3.1596, 2.7386, 2.4353, 1.9169, 1.4116, 1.7452], + device='cuda:0'), covar=tensor([0.0752, 0.0825, 0.0129, 0.0212, 0.0340, 0.0368, 0.0528, 0.0652], + device='cuda:0'), in_proj_covar=tensor([0.0471, 0.0387, 0.0282, 0.0311, 0.0413, 0.0355, 0.0373, 0.0408], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 20:11:35,936 INFO [optim.py:369] (0/4) 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 20:11:39,541 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11618.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 20:11:40,716 WARNING [train.py:1067] (0/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-05 20:11:50,745 INFO [train.py:901] (0/4) Epoch 2, batch 3550, loss[loss=0.3703, simple_loss=0.4003, pruned_loss=0.1702, over 8075.00 frames. ], tot_loss[loss=0.3753, simple_loss=0.4101, pruned_loss=0.1703, over 1613573.65 frames. ], batch size: 21, lr: 3.08e-02, grad_scale: 8.0 +2023-02-05 20:11:57,562 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11643.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 20:12:02,962 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11651.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:12:04,344 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11653.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:12:11,698 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7643, 1.5586, 2.4304, 2.0885, 2.1015, 1.4760, 1.2852, 1.5560], + device='cuda:0'), covar=tensor([0.0649, 0.0544, 0.0133, 0.0173, 0.0220, 0.0343, 0.0453, 0.0404], + device='cuda:0'), in_proj_covar=tensor([0.0476, 0.0387, 0.0284, 0.0313, 0.0412, 0.0353, 0.0372, 0.0413], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 20:12:14,505 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.94 vs. limit=2.0 +2023-02-05 20:12:25,617 INFO [train.py:901] (0/4) Epoch 2, batch 3600, loss[loss=0.3544, simple_loss=0.3804, pruned_loss=0.1642, over 7231.00 frames. ], tot_loss[loss=0.3774, simple_loss=0.4115, pruned_loss=0.1716, over 1614394.58 frames. ], batch size: 16, lr: 3.08e-02, grad_scale: 8.0 +2023-02-05 20:12:29,814 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7656, 1.1803, 3.2026, 1.1876, 2.0682, 3.4983, 3.3609, 3.0813], + device='cuda:0'), covar=tensor([0.1123, 0.1801, 0.0398, 0.1954, 0.0893, 0.0240, 0.0248, 0.0511], + device='cuda:0'), in_proj_covar=tensor([0.0239, 0.0267, 0.0177, 0.0252, 0.0202, 0.0150, 0.0146, 0.0217], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 20:12:38,541 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 +2023-02-05 20:12:45,376 INFO [optim.py:369] (0/4) 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 20:12:48,303 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11717.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:12:59,433 INFO [train.py:901] (0/4) Epoch 2, batch 3650, loss[loss=0.4306, simple_loss=0.4519, pruned_loss=0.2047, over 8663.00 frames. ], tot_loss[loss=0.3768, simple_loss=0.4108, pruned_loss=0.1714, over 1612002.77 frames. ], batch size: 34, lr: 3.07e-02, grad_scale: 8.0 +2023-02-05 20:13:07,671 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.94 vs. limit=2.0 +2023-02-05 20:13:33,833 INFO [train.py:901] (0/4) Epoch 2, batch 3700, loss[loss=0.3875, simple_loss=0.4147, pruned_loss=0.1802, over 8284.00 frames. ], tot_loss[loss=0.3761, simple_loss=0.41, pruned_loss=0.1711, over 1612519.33 frames. ], batch size: 23, lr: 3.06e-02, grad_scale: 8.0 +2023-02-05 20:13:44,419 WARNING [train.py:1067] (0/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-05 20:13:53,763 INFO [optim.py:369] (0/4) 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 20:14:08,520 INFO [train.py:901] (0/4) Epoch 2, batch 3750, loss[loss=0.4344, simple_loss=0.4451, pruned_loss=0.2118, over 8341.00 frames. ], tot_loss[loss=0.3742, simple_loss=0.4087, pruned_loss=0.1699, over 1612558.19 frames. ], batch size: 49, lr: 3.06e-02, grad_scale: 8.0 +2023-02-05 20:14:08,622 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11834.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:14:28,580 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11864.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:14:29,992 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6657, 1.9331, 3.5976, 1.1910, 2.4342, 1.8834, 1.6604, 2.0237], + device='cuda:0'), covar=tensor([0.0921, 0.1204, 0.0334, 0.1606, 0.0877, 0.1305, 0.0916, 0.1275], + device='cuda:0'), in_proj_covar=tensor([0.0364, 0.0349, 0.0380, 0.0408, 0.0457, 0.0414, 0.0371, 0.0460], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:0') +2023-02-05 20:14:42,002 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 +2023-02-05 20:14:43,036 INFO [train.py:901] (0/4) Epoch 2, batch 3800, loss[loss=0.3917, simple_loss=0.418, pruned_loss=0.1827, over 7647.00 frames. ], tot_loss[loss=0.3755, simple_loss=0.4096, pruned_loss=0.1707, over 1615938.77 frames. ], batch size: 19, lr: 3.05e-02, grad_scale: 8.0 +2023-02-05 20:14:49,683 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=11894.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:14:58,752 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11907.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:15:02,632 INFO [optim.py:369] (0/4) 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 20:15:16,321 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11932.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:15:17,491 INFO [train.py:901] (0/4) Epoch 2, batch 3850, loss[loss=0.3915, simple_loss=0.4246, pruned_loss=0.1792, over 8357.00 frames. ], tot_loss[loss=0.373, simple_loss=0.4079, pruned_loss=0.1691, over 1621345.70 frames. ], batch size: 24, lr: 3.05e-02, grad_scale: 8.0 +2023-02-05 20:15:20,310 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11938.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:15:39,723 INFO [zipformer.py:1185] (0/4) 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] (0/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-05 20:15:51,654 INFO [train.py:901] (0/4) Epoch 2, batch 3900, loss[loss=0.3289, simple_loss=0.374, pruned_loss=0.1418, over 7778.00 frames. ], tot_loss[loss=0.3734, simple_loss=0.4081, pruned_loss=0.1693, over 1619684.08 frames. ], batch size: 19, lr: 3.04e-02, grad_scale: 8.0 +2023-02-05 20:15:52,513 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7568, 2.0593, 1.9535, 2.7044, 1.1974, 1.2514, 1.7002, 2.1205], + device='cuda:0'), covar=tensor([0.1518, 0.1899, 0.1404, 0.0438, 0.2371, 0.2538, 0.2244, 0.1364], + device='cuda:0'), in_proj_covar=tensor([0.0323, 0.0342, 0.0329, 0.0214, 0.0337, 0.0338, 0.0373, 0.0312], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0005, 0.0004], + device='cuda:0') +2023-02-05 20:16:01,112 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=11997.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:16:03,786 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-12000.pt +2023-02-05 20:16:06,073 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12002.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:16:10,872 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 20:16:20,221 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9630, 1.7023, 1.2848, 1.2523, 1.7191, 1.4474, 1.5936, 1.5684], + device='cuda:0'), covar=tensor([0.0818, 0.1493, 0.2202, 0.1774, 0.0727, 0.1673, 0.0946, 0.0812], + device='cuda:0'), in_proj_covar=tensor([0.0226, 0.0260, 0.0286, 0.0253, 0.0224, 0.0247, 0.0222, 0.0211], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0004], + device='cuda:0') +2023-02-05 20:16:28,141 INFO [train.py:901] (0/4) Epoch 2, batch 3950, loss[loss=0.3896, simple_loss=0.426, pruned_loss=0.1766, over 8690.00 frames. ], tot_loss[loss=0.375, simple_loss=0.4089, pruned_loss=0.1705, over 1616696.45 frames. ], batch size: 34, lr: 3.04e-02, grad_scale: 8.0 +2023-02-05 20:16:46,991 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=12061.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:17:02,473 INFO [train.py:901] (0/4) Epoch 2, batch 4000, loss[loss=0.3703, simple_loss=0.4063, pruned_loss=0.1671, over 8460.00 frames. ], tot_loss[loss=0.3724, simple_loss=0.4067, pruned_loss=0.169, over 1610200.88 frames. ], batch size: 29, lr: 3.03e-02, grad_scale: 8.0 +2023-02-05 20:17:09,203 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) 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 20:17:36,866 INFO [train.py:901] (0/4) Epoch 2, batch 4050, loss[loss=0.3806, simple_loss=0.4221, pruned_loss=0.1696, over 8322.00 frames. ], tot_loss[loss=0.3732, simple_loss=0.4081, pruned_loss=0.1692, over 1614643.44 frames. ], batch size: 25, lr: 3.03e-02, grad_scale: 16.0 +2023-02-05 20:17:46,315 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0626, 3.7796, 2.3623, 2.6711, 2.4946, 2.4790, 2.7830, 2.5737], + device='cuda:0'), covar=tensor([0.1776, 0.0426, 0.0953, 0.1064, 0.1144, 0.0996, 0.1338, 0.1254], + device='cuda:0'), in_proj_covar=tensor([0.0372, 0.0251, 0.0352, 0.0313, 0.0366, 0.0321, 0.0358, 0.0329], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:0') +2023-02-05 20:17:48,188 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([0.9965, 1.0422, 1.0604, 0.8912, 0.6842, 0.9900, 0.1500, 0.6318], + device='cuda:0'), covar=tensor([0.0747, 0.0707, 0.0486, 0.0757, 0.1069, 0.0452, 0.2092, 0.1121], + device='cuda:0'), in_proj_covar=tensor([0.0116, 0.0102, 0.0090, 0.0143, 0.0123, 0.0085, 0.0157, 0.0125], + device='cuda:0'), out_proj_covar=tensor([1.1421e-04, 1.0727e-04, 8.9514e-05, 1.3675e-04, 1.2681e-04, 8.4791e-05, + 1.5188e-04, 1.2796e-04], device='cuda:0') +2023-02-05 20:18:06,022 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12176.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:18:06,225 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.79 vs. limit=2.0 +2023-02-05 20:18:07,290 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=12178.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:18:11,165 INFO [train.py:901] (0/4) Epoch 2, batch 4100, loss[loss=0.3238, simple_loss=0.3689, pruned_loss=0.1393, over 7806.00 frames. ], tot_loss[loss=0.3729, simple_loss=0.4074, pruned_loss=0.1692, over 1614059.59 frames. ], batch size: 19, lr: 3.02e-02, grad_scale: 16.0 +2023-02-05 20:18:27,596 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=12208.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:18:30,899 INFO [optim.py:369] (0/4) 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 20:18:47,040 INFO [train.py:901] (0/4) Epoch 2, batch 4150, loss[loss=0.3979, simple_loss=0.4167, pruned_loss=0.1895, over 8564.00 frames. ], tot_loss[loss=0.3719, simple_loss=0.4067, pruned_loss=0.1686, over 1612815.74 frames. ], batch size: 31, lr: 3.02e-02, grad_scale: 16.0 +2023-02-05 20:19:08,901 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12265.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:19:20,420 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=12282.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:19:21,722 INFO [train.py:901] (0/4) Epoch 2, batch 4200, loss[loss=0.3241, simple_loss=0.366, pruned_loss=0.1412, over 8132.00 frames. ], tot_loss[loss=0.3732, simple_loss=0.408, pruned_loss=0.1692, over 1614372.38 frames. ], batch size: 22, lr: 3.01e-02, grad_scale: 16.0 +2023-02-05 20:19:25,930 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12290.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:19:28,617 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12293.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:19:40,048 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=12310.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:19:42,058 INFO [optim.py:369] (0/4) 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 20:19:43,524 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-05 20:19:49,252 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12323.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:19:57,052 INFO [train.py:901] (0/4) Epoch 2, batch 4250, loss[loss=0.3528, simple_loss=0.3956, pruned_loss=0.155, over 8201.00 frames. ], tot_loss[loss=0.3741, simple_loss=0.4088, pruned_loss=0.1697, over 1611205.27 frames. ], batch size: 23, lr: 3.01e-02, grad_scale: 16.0 +2023-02-05 20:20:00,212 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 +2023-02-05 20:20:06,019 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=12346.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:20:06,650 WARNING [train.py:1067] (0/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-05 20:20:20,971 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12368.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:20:31,220 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.87 vs. limit=2.0 +2023-02-05 20:20:32,224 INFO [train.py:901] (0/4) Epoch 2, batch 4300, loss[loss=0.3655, simple_loss=0.4131, pruned_loss=0.159, over 8348.00 frames. ], tot_loss[loss=0.3727, simple_loss=0.4076, pruned_loss=0.1689, over 1609601.53 frames. ], batch size: 26, lr: 3.00e-02, grad_scale: 16.0 +2023-02-05 20:20:38,551 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12393.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:20:41,197 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12397.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:20:53,213 INFO [optim.py:369] (0/4) 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 20:21:00,840 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12432.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:21:06,899 INFO [train.py:901] (0/4) Epoch 2, batch 4350, loss[loss=0.4486, simple_loss=0.4523, pruned_loss=0.2224, over 8345.00 frames. ], tot_loss[loss=0.3732, simple_loss=0.4086, pruned_loss=0.1689, over 1615219.28 frames. ], batch size: 25, lr: 2.99e-02, grad_scale: 8.0 +2023-02-05 20:21:09,738 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=12438.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:21:23,500 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12457.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:21:26,031 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12461.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:21:27,749 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 +2023-02-05 20:21:28,045 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12464.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:21:37,547 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.9319, 4.2014, 3.6592, 1.7917, 3.5663, 3.4660, 3.7527, 3.0309], + device='cuda:0'), covar=tensor([0.0849, 0.0402, 0.0782, 0.3533, 0.0487, 0.0517, 0.0904, 0.0630], + device='cuda:0'), in_proj_covar=tensor([0.0319, 0.0217, 0.0254, 0.0341, 0.0231, 0.0181, 0.0238, 0.0168], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:0') +2023-02-05 20:21:38,137 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 2, batch 4400, loss[loss=0.347, simple_loss=0.3845, pruned_loss=0.1548, over 7802.00 frames. ], tot_loss[loss=0.3726, simple_loss=0.4082, pruned_loss=0.1685, over 1616415.41 frames. ], batch size: 20, lr: 2.99e-02, grad_scale: 8.0 +2023-02-05 20:22:02,392 INFO [optim.py:369] (0/4) 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 20:22:16,721 INFO [train.py:901] (0/4) Epoch 2, batch 4450, loss[loss=0.3578, simple_loss=0.4107, pruned_loss=0.1525, over 8507.00 frames. ], tot_loss[loss=0.3713, simple_loss=0.4076, pruned_loss=0.1675, over 1616637.45 frames. ], batch size: 28, lr: 2.98e-02, grad_scale: 8.0 +2023-02-05 20:22:17,400 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-05 20:22:27,259 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12549.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:22:29,906 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12553.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:22:44,461 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8545, 2.3552, 2.7091, 1.7560, 1.5438, 2.6713, 0.4478, 1.7382], + device='cuda:0'), covar=tensor([0.1350, 0.1099, 0.0492, 0.0787, 0.1487, 0.0323, 0.2597, 0.1027], + device='cuda:0'), in_proj_covar=tensor([0.0120, 0.0099, 0.0090, 0.0149, 0.0130, 0.0082, 0.0161, 0.0123], + device='cuda:0'), out_proj_covar=tensor([1.2038e-04, 1.0631e-04, 9.2233e-05, 1.4435e-04, 1.3459e-04, 8.3272e-05, + 1.5650e-04, 1.2923e-04], device='cuda:0') +2023-02-05 20:22:45,076 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12574.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:22:49,129 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12579.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:22:52,403 INFO [train.py:901] (0/4) Epoch 2, batch 4500, loss[loss=0.3433, simple_loss=0.3728, pruned_loss=0.1569, over 7796.00 frames. ], tot_loss[loss=0.3715, simple_loss=0.4082, pruned_loss=0.1673, over 1615691.68 frames. ], batch size: 19, lr: 2.98e-02, grad_scale: 8.0 +2023-02-05 20:22:52,541 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.8658, 1.1474, 5.6198, 2.3376, 5.0398, 4.8295, 5.3162, 5.2288], + device='cuda:0'), covar=tensor([0.0227, 0.3337, 0.0183, 0.1311, 0.0678, 0.0271, 0.0221, 0.0248], + device='cuda:0'), in_proj_covar=tensor([0.0182, 0.0376, 0.0234, 0.0268, 0.0323, 0.0258, 0.0241, 0.0261], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001], + device='cuda:0') +2023-02-05 20:23:06,034 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12604.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:23:12,542 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-05 20:23:13,219 INFO [optim.py:369] (0/4) 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 20:23:18,779 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.65 vs. limit=2.0 +2023-02-05 20:23:27,081 INFO [train.py:901] (0/4) Epoch 2, batch 4550, loss[loss=0.3634, simple_loss=0.3845, pruned_loss=0.1711, over 7534.00 frames. ], tot_loss[loss=0.3703, simple_loss=0.4072, pruned_loss=0.1668, over 1614997.52 frames. ], batch size: 18, lr: 2.97e-02, grad_scale: 8.0 +2023-02-05 20:23:40,657 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12653.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:23:43,559 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.98 vs. limit=5.0 +2023-02-05 20:23:57,693 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12678.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:23:59,752 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12681.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:24:01,601 INFO [train.py:901] (0/4) Epoch 2, batch 4600, loss[loss=0.4312, simple_loss=0.4478, pruned_loss=0.2073, over 8475.00 frames. ], tot_loss[loss=0.372, simple_loss=0.4078, pruned_loss=0.1681, over 1613704.52 frames. ], batch size: 49, lr: 2.97e-02, grad_scale: 8.0 +2023-02-05 20:24:17,913 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12706.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:24:23,153 INFO [optim.py:369] (0/4) 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 20:24:25,430 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12717.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:24:26,767 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2179, 1.5786, 1.8903, 1.5102, 1.0029, 1.9735, 0.2560, 0.9874], + device='cuda:0'), covar=tensor([0.1036, 0.0689, 0.0653, 0.0701, 0.1056, 0.0464, 0.2268, 0.0946], + device='cuda:0'), in_proj_covar=tensor([0.0110, 0.0096, 0.0086, 0.0145, 0.0121, 0.0081, 0.0154, 0.0120], + device='cuda:0'), out_proj_covar=tensor([1.1134e-04, 1.0328e-04, 8.8480e-05, 1.4119e-04, 1.2655e-04, 8.2358e-05, + 1.5142e-04, 1.2675e-04], device='cuda:0') +2023-02-05 20:24:37,083 INFO [train.py:901] (0/4) Epoch 2, batch 4650, loss[loss=0.3435, simple_loss=0.3916, pruned_loss=0.1477, over 8128.00 frames. ], tot_loss[loss=0.3731, simple_loss=0.4087, pruned_loss=0.1688, over 1617402.42 frames. ], batch size: 22, lr: 2.96e-02, grad_scale: 8.0 +2023-02-05 20:24:42,614 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12742.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:24:52,923 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.14 vs. limit=5.0 +2023-02-05 20:25:08,378 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4572, 2.4756, 1.3374, 2.1336, 2.1265, 1.1442, 1.6881, 2.0888], + device='cuda:0'), covar=tensor([0.1558, 0.0480, 0.1738, 0.0892, 0.1182, 0.1732, 0.1518, 0.0838], + device='cuda:0'), in_proj_covar=tensor([0.0368, 0.0249, 0.0359, 0.0309, 0.0360, 0.0323, 0.0358, 0.0323], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], + device='cuda:0') +2023-02-05 20:25:09,733 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12781.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:25:11,643 INFO [train.py:901] (0/4) Epoch 2, batch 4700, loss[loss=0.4061, simple_loss=0.4392, pruned_loss=0.1865, over 8308.00 frames. ], tot_loss[loss=0.3756, simple_loss=0.4104, pruned_loss=0.1704, over 1617183.72 frames. ], batch size: 49, lr: 2.96e-02, grad_scale: 8.0 +2023-02-05 20:25:28,064 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=12808.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:25:29,569 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12809.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:25:32,793 INFO [optim.py:369] (0/4) 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 20:25:47,179 INFO [train.py:901] (0/4) Epoch 2, batch 4750, loss[loss=0.4072, simple_loss=0.4357, pruned_loss=0.1894, over 8097.00 frames. ], tot_loss[loss=0.3739, simple_loss=0.4083, pruned_loss=0.1697, over 1614787.54 frames. ], batch size: 23, lr: 2.95e-02, grad_scale: 8.0 +2023-02-05 20:25:47,402 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12834.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:26:09,928 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2196, 4.0311, 2.4821, 3.1473, 3.5450, 2.3095, 2.5758, 3.1878], + device='cuda:0'), covar=tensor([0.1346, 0.0461, 0.1036, 0.0724, 0.0605, 0.1089, 0.1172, 0.0790], + device='cuda:0'), in_proj_covar=tensor([0.0376, 0.0251, 0.0363, 0.0318, 0.0373, 0.0326, 0.0371, 0.0332], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-05 20:26:18,690 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-05 20:26:20,739 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-05 20:26:22,726 INFO [train.py:901] (0/4) Epoch 2, batch 4800, loss[loss=0.4267, simple_loss=0.444, pruned_loss=0.2047, over 8199.00 frames. ], tot_loss[loss=0.3731, simple_loss=0.4082, pruned_loss=0.169, over 1612854.36 frames. ], batch size: 23, lr: 2.95e-02, grad_scale: 8.0 +2023-02-05 20:26:36,103 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.1286, 1.5002, 4.1889, 2.1281, 3.6102, 3.4336, 3.7083, 3.7147], + device='cuda:0'), covar=tensor([0.0305, 0.3374, 0.0237, 0.1480, 0.0839, 0.0417, 0.0354, 0.0386], + device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0387, 0.0238, 0.0275, 0.0332, 0.0264, 0.0250, 0.0268], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001], + device='cuda:0') +2023-02-05 20:26:43,398 INFO [optim.py:369] (0/4) 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 20:26:49,680 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 2, batch 4850, loss[loss=0.4242, simple_loss=0.4375, pruned_loss=0.2055, over 8358.00 frames. ], tot_loss[loss=0.3707, simple_loss=0.4066, pruned_loss=0.1674, over 1610731.86 frames. ], batch size: 24, lr: 2.94e-02, grad_scale: 8.0 +2023-02-05 20:27:12,744 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-05 20:27:13,738 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.92 vs. limit=5.0 +2023-02-05 20:27:25,497 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.9109, 2.2977, 2.1154, 1.9106, 2.2873, 2.2236, 2.6985, 2.3688], + device='cuda:0'), covar=tensor([0.0600, 0.0986, 0.1338, 0.1283, 0.0697, 0.1160, 0.0688, 0.0575], + device='cuda:0'), in_proj_covar=tensor([0.0223, 0.0255, 0.0276, 0.0248, 0.0225, 0.0243, 0.0214, 0.0211], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0004], + device='cuda:0') +2023-02-05 20:27:32,103 INFO [train.py:901] (0/4) Epoch 2, batch 4900, loss[loss=0.4049, simple_loss=0.4232, pruned_loss=0.1933, over 7690.00 frames. ], tot_loss[loss=0.3704, simple_loss=0.4061, pruned_loss=0.1674, over 1610313.74 frames. ], batch size: 18, lr: 2.94e-02, grad_scale: 8.0 +2023-02-05 20:27:53,284 INFO [optim.py:369] (0/4) 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 20:28:06,715 INFO [train.py:901] (0/4) Epoch 2, batch 4950, loss[loss=0.373, simple_loss=0.3897, pruned_loss=0.1782, over 7274.00 frames. ], tot_loss[loss=0.3716, simple_loss=0.4066, pruned_loss=0.1683, over 1610575.29 frames. ], batch size: 16, lr: 2.93e-02, grad_scale: 8.0 +2023-02-05 20:28:41,854 INFO [train.py:901] (0/4) Epoch 2, batch 5000, loss[loss=0.2924, simple_loss=0.3592, pruned_loss=0.1128, over 8612.00 frames. ], tot_loss[loss=0.3715, simple_loss=0.4067, pruned_loss=0.1682, over 1609350.38 frames. ], batch size: 39, lr: 2.93e-02, grad_scale: 8.0 +2023-02-05 20:29:02,462 INFO [optim.py:369] (0/4) 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 20:29:05,924 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7914, 1.8728, 3.2703, 1.1002, 2.3638, 1.8729, 1.6126, 2.0113], + device='cuda:0'), covar=tensor([0.0780, 0.1166, 0.0307, 0.1689, 0.0897, 0.1319, 0.0800, 0.1215], + device='cuda:0'), in_proj_covar=tensor([0.0385, 0.0368, 0.0407, 0.0443, 0.0491, 0.0432, 0.0383, 0.0491], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:0') +2023-02-05 20:29:09,710 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 2, batch 5050, loss[loss=0.3176, simple_loss=0.3674, pruned_loss=0.1339, over 8468.00 frames. ], tot_loss[loss=0.3706, simple_loss=0.4061, pruned_loss=0.1675, over 1611000.11 frames. ], batch size: 49, lr: 2.92e-02, grad_scale: 4.0 +2023-02-05 20:29:47,469 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=13179.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:29:47,963 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-05 20:29:50,615 INFO [train.py:901] (0/4) Epoch 2, batch 5100, loss[loss=0.3599, simple_loss=0.3968, pruned_loss=0.1615, over 8185.00 frames. ], tot_loss[loss=0.3696, simple_loss=0.4056, pruned_loss=0.1667, over 1610247.05 frames. ], batch size: 23, lr: 2.92e-02, grad_scale: 4.0 +2023-02-05 20:30:04,619 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=13204.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:30:08,423 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8305, 2.6706, 1.9738, 3.0443, 1.4531, 1.2380, 1.7472, 2.4174], + device='cuda:0'), covar=tensor([0.1281, 0.0998, 0.1579, 0.0401, 0.1750, 0.2094, 0.1947, 0.0919], + device='cuda:0'), in_proj_covar=tensor([0.0314, 0.0337, 0.0327, 0.0231, 0.0320, 0.0334, 0.0373, 0.0309], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004, 0.0005, 0.0004], + device='cuda:0') +2023-02-05 20:30:11,532 INFO [optim.py:369] (0/4) 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 20:30:24,591 INFO [train.py:901] (0/4) Epoch 2, batch 5150, loss[loss=0.3651, simple_loss=0.4141, pruned_loss=0.158, over 8199.00 frames. ], tot_loss[loss=0.3701, simple_loss=0.4063, pruned_loss=0.167, over 1608583.38 frames. ], batch size: 23, lr: 2.91e-02, grad_scale: 4.0 +2023-02-05 20:30:28,694 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=13240.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:30:58,999 INFO [train.py:901] (0/4) Epoch 2, batch 5200, loss[loss=0.3956, simple_loss=0.4315, pruned_loss=0.1798, over 8500.00 frames. ], tot_loss[loss=0.3694, simple_loss=0.4057, pruned_loss=0.1665, over 1607894.94 frames. ], batch size: 29, lr: 2.91e-02, grad_scale: 8.0 +2023-02-05 20:31:02,045 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.64 vs. limit=5.0 +2023-02-05 20:31:03,170 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.0407, 1.2510, 4.2910, 1.7413, 3.7251, 3.6479, 3.6564, 3.7757], + device='cuda:0'), covar=tensor([0.0402, 0.3022, 0.0262, 0.1464, 0.0844, 0.0379, 0.0340, 0.0381], + device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0380, 0.0241, 0.0272, 0.0332, 0.0260, 0.0242, 0.0264], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001], + device='cuda:0') +2023-02-05 20:31:05,932 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2994, 2.4133, 1.5157, 2.0807, 1.8979, 1.3053, 1.4889, 2.2304], + device='cuda:0'), covar=tensor([0.1233, 0.0410, 0.1015, 0.0636, 0.0798, 0.1116, 0.1212, 0.0658], + device='cuda:0'), in_proj_covar=tensor([0.0379, 0.0249, 0.0367, 0.0317, 0.0366, 0.0336, 0.0375, 0.0342], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-05 20:31:20,904 INFO [optim.py:369] (0/4) 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 20:31:26,493 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2695, 1.4839, 1.5822, 1.2620, 1.0390, 1.5130, 0.1484, 0.5467], + device='cuda:0'), covar=tensor([0.0976, 0.0665, 0.0451, 0.0649, 0.1222, 0.0473, 0.1915, 0.1342], + device='cuda:0'), in_proj_covar=tensor([0.0112, 0.0094, 0.0079, 0.0135, 0.0128, 0.0085, 0.0149, 0.0121], + device='cuda:0'), out_proj_covar=tensor([1.1541e-04, 1.0473e-04, 8.3619e-05, 1.3648e-04, 1.3472e-04, 8.8883e-05, + 1.5138e-04, 1.2869e-04], device='cuda:0') +2023-02-05 20:31:33,608 INFO [train.py:901] (0/4) Epoch 2, batch 5250, loss[loss=0.3381, simple_loss=0.3961, pruned_loss=0.1401, over 8027.00 frames. ], tot_loss[loss=0.3696, simple_loss=0.4056, pruned_loss=0.1667, over 1605268.55 frames. ], batch size: 22, lr: 2.91e-02, grad_scale: 8.0 +2023-02-05 20:31:42,975 WARNING [train.py:1067] (0/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-05 20:32:07,576 INFO [train.py:901] (0/4) Epoch 2, batch 5300, loss[loss=0.3385, simple_loss=0.3808, pruned_loss=0.1482, over 8241.00 frames. ], tot_loss[loss=0.3681, simple_loss=0.4047, pruned_loss=0.1658, over 1610285.32 frames. ], batch size: 22, lr: 2.90e-02, grad_scale: 8.0 +2023-02-05 20:32:29,093 INFO [optim.py:369] (0/4) 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 20:32:42,514 INFO [train.py:901] (0/4) Epoch 2, batch 5350, loss[loss=0.332, simple_loss=0.3764, pruned_loss=0.1438, over 7522.00 frames. ], tot_loss[loss=0.3711, simple_loss=0.4068, pruned_loss=0.1677, over 1613575.61 frames. ], batch size: 18, lr: 2.90e-02, grad_scale: 8.0 +2023-02-05 20:33:16,587 INFO [train.py:901] (0/4) Epoch 2, batch 5400, loss[loss=0.2785, simple_loss=0.3217, pruned_loss=0.1176, over 7418.00 frames. ], tot_loss[loss=0.3704, simple_loss=0.406, pruned_loss=0.1674, over 1615662.08 frames. ], batch size: 17, lr: 2.89e-02, grad_scale: 8.0 +2023-02-05 20:33:24,787 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.6220, 3.7055, 3.2094, 1.6233, 3.0731, 3.0956, 3.3051, 2.7875], + device='cuda:0'), covar=tensor([0.1184, 0.0609, 0.1012, 0.4253, 0.0717, 0.0738, 0.1124, 0.0663], + device='cuda:0'), in_proj_covar=tensor([0.0314, 0.0214, 0.0255, 0.0337, 0.0231, 0.0177, 0.0237, 0.0159], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:0') +2023-02-05 20:33:24,907 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=13496.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:33:38,016 INFO [optim.py:369] (0/4) 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 20:33:41,737 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4448, 1.7906, 3.5030, 0.8976, 2.1670, 1.5632, 1.3611, 1.6985], + device='cuda:0'), covar=tensor([0.1290, 0.1499, 0.0379, 0.2224, 0.1311, 0.2026, 0.1280, 0.1804], + device='cuda:0'), in_proj_covar=tensor([0.0399, 0.0380, 0.0412, 0.0446, 0.0506, 0.0441, 0.0396, 0.0502], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:0') +2023-02-05 20:33:43,022 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=13521.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:33:51,323 INFO [train.py:901] (0/4) Epoch 2, batch 5450, loss[loss=0.4097, simple_loss=0.4338, pruned_loss=0.1928, over 8656.00 frames. ], tot_loss[loss=0.3706, simple_loss=0.4064, pruned_loss=0.1674, over 1613040.40 frames. ], batch size: 34, lr: 2.89e-02, grad_scale: 8.0 +2023-02-05 20:34:25,970 INFO [train.py:901] (0/4) Epoch 2, batch 5500, loss[loss=0.4009, simple_loss=0.4423, pruned_loss=0.1797, over 8247.00 frames. ], tot_loss[loss=0.3706, simple_loss=0.4062, pruned_loss=0.1675, over 1611775.44 frames. ], batch size: 24, lr: 2.88e-02, grad_scale: 8.0 +2023-02-05 20:34:28,057 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-05 20:34:40,937 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2759, 1.5602, 1.9620, 1.2553, 0.9590, 1.7532, 0.3123, 0.9729], + device='cuda:0'), covar=tensor([0.1326, 0.0807, 0.0515, 0.1048, 0.1579, 0.0638, 0.2625, 0.1386], + device='cuda:0'), in_proj_covar=tensor([0.0107, 0.0092, 0.0081, 0.0133, 0.0132, 0.0081, 0.0147, 0.0117], + device='cuda:0'), out_proj_covar=tensor([1.1187e-04, 1.0181e-04, 8.5565e-05, 1.3545e-04, 1.3969e-04, 8.5669e-05, + 1.5146e-04, 1.2653e-04], device='cuda:0') +2023-02-05 20:34:46,540 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 2, batch 5550, loss[loss=0.3676, simple_loss=0.3922, pruned_loss=0.1715, over 7922.00 frames. ], tot_loss[loss=0.3685, simple_loss=0.4049, pruned_loss=0.166, over 1612453.12 frames. ], batch size: 20, lr: 2.88e-02, grad_scale: 8.0 +2023-02-05 20:35:01,428 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2448, 1.4280, 1.7115, 1.3099, 1.0604, 1.7149, 0.4985, 0.9613], + device='cuda:0'), covar=tensor([0.1094, 0.1423, 0.0421, 0.0952, 0.1208, 0.0436, 0.2888, 0.1589], + device='cuda:0'), in_proj_covar=tensor([0.0113, 0.0095, 0.0083, 0.0139, 0.0136, 0.0084, 0.0154, 0.0124], + device='cuda:0'), out_proj_covar=tensor([1.1757e-04, 1.0561e-04, 8.8889e-05, 1.4163e-04, 1.4379e-04, 8.9130e-05, + 1.5758e-04, 1.3318e-04], device='cuda:0') +2023-02-05 20:35:35,316 INFO [train.py:901] (0/4) Epoch 2, batch 5600, loss[loss=0.3874, simple_loss=0.4305, pruned_loss=0.1722, over 8117.00 frames. ], tot_loss[loss=0.3686, simple_loss=0.4055, pruned_loss=0.1659, over 1612455.31 frames. ], batch size: 23, lr: 2.87e-02, grad_scale: 8.0 +2023-02-05 20:35:55,773 INFO [optim.py:369] (0/4) 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 20:35:58,636 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.5427, 1.8753, 1.4875, 1.3588, 1.8986, 1.7175, 1.8182, 2.0753], + device='cuda:0'), covar=tensor([0.1089, 0.1492, 0.2168, 0.1986, 0.0913, 0.1613, 0.1079, 0.0720], + device='cuda:0'), in_proj_covar=tensor([0.0222, 0.0253, 0.0279, 0.0249, 0.0225, 0.0242, 0.0215, 0.0206], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0005, 0.0004], + device='cuda:0') +2023-02-05 20:36:08,566 INFO [train.py:901] (0/4) Epoch 2, batch 5650, loss[loss=0.4257, simple_loss=0.4557, pruned_loss=0.1978, over 8497.00 frames. ], tot_loss[loss=0.3675, simple_loss=0.4049, pruned_loss=0.165, over 1615543.72 frames. ], batch size: 26, lr: 2.87e-02, grad_scale: 8.0 +2023-02-05 20:36:23,365 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=13755.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:36:34,187 WARNING [train.py:1067] (0/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-05 20:36:43,564 INFO [train.py:901] (0/4) Epoch 2, batch 5700, loss[loss=0.3413, simple_loss=0.4017, pruned_loss=0.1404, over 8571.00 frames. ], tot_loss[loss=0.3679, simple_loss=0.4057, pruned_loss=0.1651, over 1617233.49 frames. ], batch size: 34, lr: 2.86e-02, grad_scale: 8.0 +2023-02-05 20:37:01,016 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9227, 2.5146, 2.1373, 2.9762, 1.5649, 1.3056, 2.0211, 2.1799], + device='cuda:0'), covar=tensor([0.1257, 0.1266, 0.1349, 0.0429, 0.1771, 0.2256, 0.1694, 0.1368], + device='cuda:0'), in_proj_covar=tensor([0.0302, 0.0328, 0.0318, 0.0231, 0.0308, 0.0327, 0.0367, 0.0306], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004, 0.0005, 0.0004], + device='cuda:0') +2023-02-05 20:37:05,591 INFO [optim.py:369] (0/4) 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 20:37:18,916 INFO [train.py:901] (0/4) Epoch 2, batch 5750, loss[loss=0.3244, simple_loss=0.3823, pruned_loss=0.1332, over 8241.00 frames. ], tot_loss[loss=0.3693, simple_loss=0.4061, pruned_loss=0.1663, over 1615399.99 frames. ], batch size: 22, lr: 2.86e-02, grad_scale: 8.0 +2023-02-05 20:37:38,963 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-05 20:37:54,468 INFO [train.py:901] (0/4) Epoch 2, batch 5800, loss[loss=0.3548, simple_loss=0.3964, pruned_loss=0.1566, over 8137.00 frames. ], tot_loss[loss=0.3667, simple_loss=0.404, pruned_loss=0.1647, over 1613711.81 frames. ], batch size: 22, lr: 2.85e-02, grad_scale: 8.0 +2023-02-05 20:38:02,404 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.07 vs. limit=5.0 +2023-02-05 20:38:15,740 INFO [optim.py:369] (0/4) 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 20:38:29,064 INFO [train.py:901] (0/4) Epoch 2, batch 5850, loss[loss=0.3933, simple_loss=0.4201, pruned_loss=0.1832, over 7923.00 frames. ], tot_loss[loss=0.3671, simple_loss=0.4043, pruned_loss=0.165, over 1616278.07 frames. ], batch size: 20, lr: 2.85e-02, grad_scale: 8.0 +2023-02-05 20:38:39,342 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6685, 2.3928, 1.6818, 2.2351, 2.0453, 1.3622, 1.7360, 2.3449], + device='cuda:0'), covar=tensor([0.1113, 0.0383, 0.0974, 0.0666, 0.0783, 0.1277, 0.0979, 0.0615], + device='cuda:0'), in_proj_covar=tensor([0.0375, 0.0245, 0.0359, 0.0318, 0.0362, 0.0333, 0.0356, 0.0331], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], + device='cuda:0') +2023-02-05 20:39:03,917 INFO [train.py:901] (0/4) Epoch 2, batch 5900, loss[loss=0.4424, simple_loss=0.4408, pruned_loss=0.222, over 7293.00 frames. ], tot_loss[loss=0.367, simple_loss=0.404, pruned_loss=0.165, over 1609435.04 frames. ], batch size: 72, lr: 2.84e-02, grad_scale: 8.0 +2023-02-05 20:39:15,923 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-14000.pt +2023-02-05 20:39:27,076 INFO [optim.py:369] (0/4) 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 20:39:40,162 INFO [train.py:901] (0/4) Epoch 2, batch 5950, loss[loss=0.4412, simple_loss=0.4528, pruned_loss=0.2148, over 7492.00 frames. ], tot_loss[loss=0.3669, simple_loss=0.4041, pruned_loss=0.1648, over 1610885.59 frames. ], batch size: 71, lr: 2.84e-02, grad_scale: 8.0 +2023-02-05 20:40:14,633 INFO [train.py:901] (0/4) Epoch 2, batch 6000, loss[loss=0.2961, simple_loss=0.338, pruned_loss=0.1271, over 7436.00 frames. ], tot_loss[loss=0.364, simple_loss=0.4011, pruned_loss=0.1634, over 1607719.44 frames. ], batch size: 17, lr: 2.84e-02, grad_scale: 8.0 +2023-02-05 20:40:14,634 INFO [train.py:926] (0/4) Computing validation loss +2023-02-05 20:40:27,827 INFO [train.py:935] (0/4) Epoch 2, validation: loss=0.2758, simple_loss=0.3606, pruned_loss=0.0955, over 944034.00 frames. +2023-02-05 20:40:27,828 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6484MB +2023-02-05 20:40:32,051 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14090.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:40:38,735 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=14099.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:40:49,504 INFO [optim.py:369] (0/4) 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 20:40:53,745 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14121.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:41:02,696 INFO [train.py:901] (0/4) Epoch 2, batch 6050, loss[loss=0.3269, simple_loss=0.3628, pruned_loss=0.1455, over 7569.00 frames. ], tot_loss[loss=0.3635, simple_loss=0.4006, pruned_loss=0.1632, over 1605917.21 frames. ], batch size: 18, lr: 2.83e-02, grad_scale: 8.0 +2023-02-05 20:41:05,424 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14138.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 20:41:25,413 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4669, 2.6702, 1.5314, 2.0932, 2.0091, 1.2944, 1.9910, 2.0655], + device='cuda:0'), covar=tensor([0.1217, 0.0335, 0.1115, 0.0705, 0.0867, 0.1146, 0.0927, 0.0772], + device='cuda:0'), in_proj_covar=tensor([0.0370, 0.0250, 0.0360, 0.0320, 0.0358, 0.0327, 0.0349, 0.0335], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], + device='cuda:0') +2023-02-05 20:41:29,380 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.6110, 3.0805, 2.5522, 3.8594, 1.9601, 1.4959, 2.1454, 2.7582], + device='cuda:0'), covar=tensor([0.1050, 0.1341, 0.1417, 0.0293, 0.1768, 0.2287, 0.2211, 0.1123], + device='cuda:0'), in_proj_covar=tensor([0.0310, 0.0339, 0.0324, 0.0234, 0.0313, 0.0335, 0.0369, 0.0305], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0005, 0.0003, 0.0004, 0.0005, 0.0005, 0.0004], + device='cuda:0') +2023-02-05 20:41:33,311 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4620, 2.1220, 3.3062, 0.9159, 2.1461, 1.5169, 1.4961, 1.8047], + device='cuda:0'), covar=tensor([0.1152, 0.1257, 0.0459, 0.2241, 0.1242, 0.1983, 0.1099, 0.1769], + device='cuda:0'), in_proj_covar=tensor([0.0390, 0.0373, 0.0418, 0.0442, 0.0493, 0.0437, 0.0394, 0.0485], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:0') +2023-02-05 20:41:37,173 INFO [train.py:901] (0/4) Epoch 2, batch 6100, loss[loss=0.3469, simple_loss=0.3946, pruned_loss=0.1496, over 8470.00 frames. ], tot_loss[loss=0.3644, simple_loss=0.401, pruned_loss=0.1639, over 1608330.75 frames. ], batch size: 25, lr: 2.83e-02, grad_scale: 8.0 +2023-02-05 20:41:40,655 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.1124, 5.2580, 4.4812, 2.0616, 4.3949, 4.5809, 4.8540, 4.1174], + device='cuda:0'), covar=tensor([0.0525, 0.0233, 0.0676, 0.3387, 0.0340, 0.0430, 0.0526, 0.0459], + device='cuda:0'), in_proj_covar=tensor([0.0316, 0.0213, 0.0254, 0.0339, 0.0225, 0.0183, 0.0233, 0.0160], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001], + device='cuda:0') +2023-02-05 20:41:51,630 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9317, 3.8509, 2.1156, 2.0625, 2.5413, 1.7722, 2.2088, 2.4726], + device='cuda:0'), covar=tensor([0.1491, 0.0222, 0.0978, 0.1179, 0.0873, 0.1013, 0.1330, 0.0953], + device='cuda:0'), in_proj_covar=tensor([0.0372, 0.0249, 0.0358, 0.0322, 0.0356, 0.0327, 0.0355, 0.0333], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], + device='cuda:0') +2023-02-05 20:41:58,388 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14214.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:41:58,921 INFO [optim.py:369] (0/4) 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 20:42:07,558 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3212, 1.5617, 2.0475, 1.3156, 0.9524, 1.7995, 0.2355, 0.8641], + device='cuda:0'), covar=tensor([0.2206, 0.1099, 0.0493, 0.1090, 0.1612, 0.0516, 0.3100, 0.1310], + device='cuda:0'), in_proj_covar=tensor([0.0113, 0.0099, 0.0085, 0.0143, 0.0143, 0.0086, 0.0161, 0.0123], + device='cuda:0'), out_proj_covar=tensor([1.2028e-04, 1.1350e-04, 9.1997e-05, 1.4874e-04, 1.5364e-04, 9.4816e-05, + 1.6764e-04, 1.3480e-04], device='cuda:0') +2023-02-05 20:42:08,080 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-05 20:42:11,469 INFO [train.py:901] (0/4) Epoch 2, batch 6150, loss[loss=0.3848, simple_loss=0.4236, pruned_loss=0.1731, over 8425.00 frames. ], tot_loss[loss=0.3625, simple_loss=0.4004, pruned_loss=0.1623, over 1611098.71 frames. ], batch size: 29, lr: 2.82e-02, grad_scale: 8.0 +2023-02-05 20:42:46,424 INFO [train.py:901] (0/4) Epoch 2, batch 6200, loss[loss=0.3236, simple_loss=0.3722, pruned_loss=0.1375, over 8036.00 frames. ], tot_loss[loss=0.3633, simple_loss=0.4006, pruned_loss=0.163, over 1609464.69 frames. ], batch size: 22, lr: 2.82e-02, grad_scale: 8.0 +2023-02-05 20:42:56,106 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5391, 1.2937, 1.3233, 1.2259, 1.4569, 1.4067, 1.2519, 1.2952], + device='cuda:0'), covar=tensor([0.0924, 0.1475, 0.2043, 0.1696, 0.0741, 0.1574, 0.1005, 0.0765], + device='cuda:0'), in_proj_covar=tensor([0.0218, 0.0246, 0.0276, 0.0243, 0.0217, 0.0241, 0.0210, 0.0205], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0005, 0.0004], + device='cuda:0') +2023-02-05 20:43:08,134 INFO [optim.py:369] (0/4) 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 20:43:15,404 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.90 vs. limit=2.0 +2023-02-05 20:43:21,530 INFO [train.py:901] (0/4) Epoch 2, batch 6250, loss[loss=0.4214, simple_loss=0.4499, pruned_loss=0.1964, over 8665.00 frames. ], tot_loss[loss=0.3634, simple_loss=0.4006, pruned_loss=0.1631, over 1605998.93 frames. ], batch size: 34, lr: 2.81e-02, grad_scale: 8.0 +2023-02-05 20:43:55,855 INFO [train.py:901] (0/4) Epoch 2, batch 6300, loss[loss=0.3613, simple_loss=0.4044, pruned_loss=0.1591, over 8245.00 frames. ], tot_loss[loss=0.3647, simple_loss=0.4016, pruned_loss=0.1639, over 1611159.72 frames. ], batch size: 24, lr: 2.81e-02, grad_scale: 8.0 +2023-02-05 20:44:17,496 INFO [optim.py:369] (0/4) 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 20:44:28,377 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14431.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:44:30,244 INFO [train.py:901] (0/4) Epoch 2, batch 6350, loss[loss=0.3815, simple_loss=0.4183, pruned_loss=0.1723, over 8246.00 frames. ], tot_loss[loss=0.3656, simple_loss=0.4022, pruned_loss=0.1645, over 1610297.35 frames. ], batch size: 24, lr: 2.81e-02, grad_scale: 8.0 +2023-02-05 20:44:30,312 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=14434.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:44:39,683 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5613, 1.8445, 3.2184, 1.0084, 2.3168, 1.6090, 1.5280, 1.9820], + device='cuda:0'), covar=tensor([0.0892, 0.1140, 0.0268, 0.1804, 0.0897, 0.1546, 0.0847, 0.1239], + device='cuda:0'), in_proj_covar=tensor([0.0407, 0.0382, 0.0432, 0.0456, 0.0510, 0.0450, 0.0402, 0.0507], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:0') +2023-02-05 20:44:51,460 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=14465.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:44:54,880 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14470.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:45:03,089 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=14482.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 20:45:04,274 INFO [train.py:901] (0/4) Epoch 2, batch 6400, loss[loss=0.3266, simple_loss=0.3781, pruned_loss=0.1376, over 8506.00 frames. ], tot_loss[loss=0.3642, simple_loss=0.4011, pruned_loss=0.1636, over 1612929.84 frames. ], batch size: 28, lr: 2.80e-02, grad_scale: 8.0 +2023-02-05 20:45:08,457 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.7873, 3.4258, 2.4943, 4.0503, 1.7635, 2.0100, 2.3725, 3.4763], + device='cuda:0'), covar=tensor([0.0879, 0.1001, 0.1186, 0.0222, 0.1768, 0.1829, 0.2034, 0.0752], + device='cuda:0'), in_proj_covar=tensor([0.0303, 0.0336, 0.0324, 0.0234, 0.0316, 0.0333, 0.0368, 0.0298], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0005, 0.0005, 0.0003, 0.0005, 0.0005, 0.0005, 0.0004], + device='cuda:0') +2023-02-05 20:45:12,409 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) 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 20:45:38,722 INFO [train.py:901] (0/4) Epoch 2, batch 6450, loss[loss=0.4155, simple_loss=0.4376, pruned_loss=0.1967, over 8134.00 frames. ], tot_loss[loss=0.3652, simple_loss=0.4019, pruned_loss=0.1642, over 1612635.77 frames. ], batch size: 22, lr: 2.80e-02, grad_scale: 8.0 +2023-02-05 20:45:48,966 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14549.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:46:10,585 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14580.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:46:13,173 INFO [train.py:901] (0/4) Epoch 2, batch 6500, loss[loss=0.306, simple_loss=0.3747, pruned_loss=0.1187, over 8367.00 frames. ], tot_loss[loss=0.3631, simple_loss=0.4007, pruned_loss=0.1628, over 1608384.69 frames. ], batch size: 24, lr: 2.79e-02, grad_scale: 8.0 +2023-02-05 20:46:22,643 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14597.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 20:46:35,358 INFO [optim.py:369] (0/4) 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 20:46:48,444 INFO [train.py:901] (0/4) Epoch 2, batch 6550, loss[loss=0.3785, simple_loss=0.42, pruned_loss=0.1685, over 8635.00 frames. ], tot_loss[loss=0.367, simple_loss=0.4039, pruned_loss=0.1651, over 1612768.45 frames. ], batch size: 34, lr: 2.79e-02, grad_scale: 8.0 +2023-02-05 20:47:16,641 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-05 20:47:23,564 INFO [train.py:901] (0/4) Epoch 2, batch 6600, loss[loss=0.3322, simple_loss=0.3905, pruned_loss=0.1369, over 8141.00 frames. ], tot_loss[loss=0.3624, simple_loss=0.4003, pruned_loss=0.1622, over 1611987.65 frames. ], batch size: 22, lr: 2.78e-02, grad_scale: 8.0 +2023-02-05 20:47:32,650 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3222, 2.0195, 1.9410, 0.4932, 1.9503, 1.3927, 0.3328, 1.7422], + device='cuda:0'), covar=tensor([0.0150, 0.0072, 0.0081, 0.0179, 0.0104, 0.0252, 0.0267, 0.0089], + device='cuda:0'), in_proj_covar=tensor([0.0194, 0.0137, 0.0124, 0.0186, 0.0135, 0.0250, 0.0199, 0.0179], + device='cuda:0'), out_proj_covar=tensor([1.1092e-04, 7.7753e-05, 7.3749e-05, 1.0455e-04, 8.1603e-05, 1.5364e-04, + 1.1549e-04, 1.0381e-04], device='cuda:0') +2023-02-05 20:47:36,575 WARNING [train.py:1067] (0/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-05 20:47:45,898 INFO [optim.py:369] (0/4) 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 20:47:58,960 INFO [train.py:901] (0/4) Epoch 2, batch 6650, loss[loss=0.3344, simple_loss=0.39, pruned_loss=0.1394, over 8353.00 frames. ], tot_loss[loss=0.3626, simple_loss=0.4008, pruned_loss=0.1622, over 1616256.76 frames. ], batch size: 24, lr: 2.78e-02, grad_scale: 8.0 +2023-02-05 20:48:16,415 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14758.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:48:25,755 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 +2023-02-05 20:48:28,656 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=14775.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:48:34,798 INFO [train.py:901] (0/4) Epoch 2, batch 6700, loss[loss=0.3493, simple_loss=0.4021, pruned_loss=0.1482, over 8528.00 frames. ], tot_loss[loss=0.3595, simple_loss=0.3981, pruned_loss=0.1604, over 1610648.67 frames. ], batch size: 28, lr: 2.78e-02, grad_scale: 8.0 +2023-02-05 20:48:50,211 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14805.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:48:56,685 INFO [optim.py:369] (0/4) 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 20:49:07,160 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14830.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:49:10,368 INFO [train.py:901] (0/4) Epoch 2, batch 6750, loss[loss=0.3697, simple_loss=0.402, pruned_loss=0.1687, over 8032.00 frames. ], tot_loss[loss=0.3608, simple_loss=0.3995, pruned_loss=0.161, over 1616820.84 frames. ], batch size: 22, lr: 2.77e-02, grad_scale: 8.0 +2023-02-05 20:49:11,940 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14836.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:49:21,258 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=14849.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:49:24,100 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14853.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 20:49:29,662 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14861.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:49:41,505 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14878.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 20:49:45,993 INFO [train.py:901] (0/4) Epoch 2, batch 6800, loss[loss=0.388, simple_loss=0.411, pruned_loss=0.1825, over 8624.00 frames. ], tot_loss[loss=0.3631, simple_loss=0.4015, pruned_loss=0.1623, over 1617587.88 frames. ], batch size: 31, lr: 2.77e-02, grad_scale: 8.0 +2023-02-05 20:49:50,353 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14890.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:49:51,039 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.5827, 3.8847, 2.3483, 2.6313, 2.8837, 1.9339, 2.1622, 2.7417], + device='cuda:0'), covar=tensor([0.1458, 0.0461, 0.0953, 0.0834, 0.0798, 0.1182, 0.1424, 0.1120], + device='cuda:0'), in_proj_covar=tensor([0.0382, 0.0247, 0.0366, 0.0317, 0.0354, 0.0338, 0.0359, 0.0332], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-05 20:49:54,312 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-05 20:49:54,456 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.9033, 0.9213, 4.0311, 1.5576, 3.4192, 3.3147, 3.5077, 3.5878], + device='cuda:0'), covar=tensor([0.0361, 0.3442, 0.0334, 0.1858, 0.0968, 0.0450, 0.0404, 0.0422], + device='cuda:0'), in_proj_covar=tensor([0.0202, 0.0381, 0.0245, 0.0280, 0.0338, 0.0267, 0.0260, 0.0275], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001], + device='cuda:0') +2023-02-05 20:50:07,691 INFO [optim.py:369] (0/4) 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 20:50:21,330 INFO [train.py:901] (0/4) Epoch 2, batch 6850, loss[loss=0.312, simple_loss=0.3833, pruned_loss=0.1204, over 8474.00 frames. ], tot_loss[loss=0.3636, simple_loss=0.402, pruned_loss=0.1626, over 1613195.98 frames. ], batch size: 25, lr: 2.76e-02, grad_scale: 8.0 +2023-02-05 20:50:42,744 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14964.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:50:45,353 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-05 20:50:57,108 INFO [train.py:901] (0/4) Epoch 2, batch 6900, loss[loss=0.3758, simple_loss=0.4174, pruned_loss=0.1671, over 8466.00 frames. ], tot_loss[loss=0.3638, simple_loss=0.4024, pruned_loss=0.1626, over 1615612.76 frames. ], batch size: 27, lr: 2.76e-02, grad_scale: 8.0 +2023-02-05 20:51:19,286 INFO [optim.py:369] (0/4) 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 20:51:20,411 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 +2023-02-05 20:51:30,020 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4509, 1.7702, 2.0734, 1.6215, 1.0669, 1.8655, 0.4272, 1.2116], + device='cuda:0'), covar=tensor([0.1515, 0.1303, 0.0687, 0.1253, 0.2607, 0.0777, 0.3898, 0.1541], + device='cuda:0'), in_proj_covar=tensor([0.0107, 0.0093, 0.0079, 0.0139, 0.0140, 0.0082, 0.0156, 0.0112], + device='cuda:0'), out_proj_covar=tensor([1.1922e-04, 1.1052e-04, 8.8005e-05, 1.4862e-04, 1.5254e-04, 9.3523e-05, + 1.6629e-04, 1.2750e-04], device='cuda:0') +2023-02-05 20:51:32,598 INFO [train.py:901] (0/4) Epoch 2, batch 6950, loss[loss=0.3712, simple_loss=0.4227, pruned_loss=0.1598, over 8591.00 frames. ], tot_loss[loss=0.3615, simple_loss=0.4004, pruned_loss=0.1614, over 1612843.20 frames. ], batch size: 34, lr: 2.75e-02, grad_scale: 8.0 +2023-02-05 20:51:46,956 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.5422, 1.6314, 5.4633, 2.1864, 4.8075, 4.5030, 4.8514, 4.7878], + device='cuda:0'), covar=tensor([0.0290, 0.2972, 0.0168, 0.1485, 0.0824, 0.0318, 0.0358, 0.0382], + device='cuda:0'), in_proj_covar=tensor([0.0208, 0.0382, 0.0244, 0.0282, 0.0341, 0.0277, 0.0261, 0.0279], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001], + device='cuda:0') +2023-02-05 20:51:56,478 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-05 20:52:05,177 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0691, 1.7345, 1.7469, 1.1704, 1.0007, 1.5230, 0.1739, 0.8051], + device='cuda:0'), covar=tensor([0.1985, 0.0935, 0.0512, 0.1106, 0.2180, 0.0536, 0.3502, 0.1464], + device='cuda:0'), in_proj_covar=tensor([0.0108, 0.0092, 0.0078, 0.0136, 0.0136, 0.0080, 0.0152, 0.0111], + device='cuda:0'), out_proj_covar=tensor([1.1993e-04, 1.0899e-04, 8.6352e-05, 1.4546e-04, 1.4939e-04, 9.1325e-05, + 1.6171e-04, 1.2568e-04], device='cuda:0') +2023-02-05 20:52:08,415 INFO [train.py:901] (0/4) Epoch 2, batch 7000, loss[loss=0.3752, simple_loss=0.4115, pruned_loss=0.1695, over 8083.00 frames. ], tot_loss[loss=0.3597, simple_loss=0.3994, pruned_loss=0.1601, over 1612175.97 frames. ], batch size: 21, lr: 2.75e-02, grad_scale: 8.0 +2023-02-05 20:52:21,512 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=15102.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:52:30,569 INFO [optim.py:369] (0/4) 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 20:52:44,346 INFO [train.py:901] (0/4) Epoch 2, batch 7050, loss[loss=0.3451, simple_loss=0.3924, pruned_loss=0.1489, over 8457.00 frames. ], tot_loss[loss=0.3613, simple_loss=0.4006, pruned_loss=0.161, over 1618706.84 frames. ], batch size: 25, lr: 2.75e-02, grad_scale: 16.0 +2023-02-05 20:52:52,913 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=15146.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:53:10,274 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 2, batch 7100, loss[loss=0.4176, simple_loss=0.4181, pruned_loss=0.2086, over 7112.00 frames. ], tot_loss[loss=0.3598, simple_loss=0.3992, pruned_loss=0.1602, over 1614933.91 frames. ], batch size: 72, lr: 2.74e-02, grad_scale: 16.0 +2023-02-05 20:53:39,773 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=15213.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:53:41,002 INFO [optim.py:369] (0/4) 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 20:53:42,524 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=15217.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:53:44,504 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=15220.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:53:52,746 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.17 vs. limit=5.0 +2023-02-05 20:53:53,627 INFO [train.py:901] (0/4) Epoch 2, batch 7150, loss[loss=0.3743, simple_loss=0.4102, pruned_loss=0.1692, over 8372.00 frames. ], tot_loss[loss=0.36, simple_loss=0.3993, pruned_loss=0.1604, over 1613879.62 frames. ], batch size: 49, lr: 2.74e-02, grad_scale: 16.0 +2023-02-05 20:54:02,063 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=15245.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:54:19,457 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 +2023-02-05 20:54:29,193 INFO [train.py:901] (0/4) Epoch 2, batch 7200, loss[loss=0.3727, simple_loss=0.4085, pruned_loss=0.1685, over 8314.00 frames. ], tot_loss[loss=0.3583, simple_loss=0.3986, pruned_loss=0.159, over 1616505.41 frames. ], batch size: 25, lr: 2.73e-02, grad_scale: 16.0 +2023-02-05 20:54:30,791 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5016, 1.7930, 1.9726, 0.3640, 2.0419, 1.3590, 0.4534, 1.7980], + device='cuda:0'), covar=tensor([0.0107, 0.0057, 0.0058, 0.0164, 0.0083, 0.0197, 0.0188, 0.0062], + device='cuda:0'), in_proj_covar=tensor([0.0197, 0.0132, 0.0120, 0.0182, 0.0135, 0.0246, 0.0198, 0.0171], + device='cuda:0'), out_proj_covar=tensor([1.1097e-04, 7.3825e-05, 6.9222e-05, 9.9501e-05, 7.9556e-05, 1.4948e-04, + 1.1297e-04, 9.7030e-05], device='cuda:0') +2023-02-05 20:54:31,471 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4264, 1.8582, 1.4614, 1.9039, 1.7247, 1.1248, 1.4738, 1.9578], + device='cuda:0'), covar=tensor([0.1219, 0.0577, 0.1198, 0.0742, 0.0879, 0.1352, 0.1134, 0.0693], + device='cuda:0'), in_proj_covar=tensor([0.0378, 0.0248, 0.0354, 0.0326, 0.0351, 0.0329, 0.0355, 0.0327], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], + device='cuda:0') +2023-02-05 20:54:51,171 INFO [optim.py:369] (0/4) 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 20:55:02,398 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8382, 1.3562, 3.2603, 1.2586, 2.1162, 3.6586, 3.4061, 3.1068], + device='cuda:0'), covar=tensor([0.1202, 0.1776, 0.0362, 0.1950, 0.0817, 0.0320, 0.0394, 0.0591], + device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0261, 0.0190, 0.0254, 0.0195, 0.0164, 0.0154, 0.0236], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 20:55:04,904 INFO [train.py:901] (0/4) Epoch 2, batch 7250, loss[loss=0.3679, simple_loss=0.4167, pruned_loss=0.1595, over 8107.00 frames. ], tot_loss[loss=0.3577, simple_loss=0.3984, pruned_loss=0.1585, over 1612649.29 frames. ], batch size: 23, lr: 2.73e-02, grad_scale: 8.0 +2023-02-05 20:55:39,921 INFO [train.py:901] (0/4) Epoch 2, batch 7300, loss[loss=0.3749, simple_loss=0.4286, pruned_loss=0.1606, over 8531.00 frames. ], tot_loss[loss=0.3564, simple_loss=0.3971, pruned_loss=0.1578, over 1610605.17 frames. ], batch size: 28, lr: 2.73e-02, grad_scale: 8.0 +2023-02-05 20:56:02,319 INFO [optim.py:369] (0/4) 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 20:56:14,885 INFO [train.py:901] (0/4) Epoch 2, batch 7350, loss[loss=0.3683, simple_loss=0.409, pruned_loss=0.1638, over 8327.00 frames. ], tot_loss[loss=0.356, simple_loss=0.3969, pruned_loss=0.1576, over 1609209.28 frames. ], batch size: 25, lr: 2.72e-02, grad_scale: 8.0 +2023-02-05 20:56:42,772 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=15473.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:56:43,901 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-05 20:56:49,847 INFO [train.py:901] (0/4) Epoch 2, batch 7400, loss[loss=0.3849, simple_loss=0.4107, pruned_loss=0.1796, over 8082.00 frames. ], tot_loss[loss=0.356, simple_loss=0.3964, pruned_loss=0.1577, over 1612713.48 frames. ], batch size: 21, lr: 2.72e-02, grad_scale: 8.0 +2023-02-05 20:56:59,507 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=15498.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:57:01,981 WARNING [train.py:1067] (0/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-05 20:57:11,815 INFO [optim.py:369] (0/4) 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 20:57:24,681 INFO [train.py:901] (0/4) Epoch 2, batch 7450, loss[loss=0.4035, simple_loss=0.4347, pruned_loss=0.1862, over 8458.00 frames. ], tot_loss[loss=0.357, simple_loss=0.397, pruned_loss=0.1585, over 1612476.92 frames. ], batch size: 27, lr: 2.71e-02, grad_scale: 8.0 +2023-02-05 20:57:40,458 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=15557.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:57:41,773 WARNING [train.py:1067] (0/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-05 20:57:59,035 INFO [train.py:901] (0/4) Epoch 2, batch 7500, loss[loss=0.298, simple_loss=0.3564, pruned_loss=0.1198, over 7811.00 frames. ], tot_loss[loss=0.3585, simple_loss=0.3975, pruned_loss=0.1597, over 1609419.37 frames. ], batch size: 20, lr: 2.71e-02, grad_scale: 8.0 +2023-02-05 20:58:21,359 INFO [optim.py:369] (0/4) 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 20:58:28,532 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 +2023-02-05 20:58:30,929 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4768, 1.5070, 1.4240, 1.3368, 2.0018, 1.6272, 1.7345, 1.7904], + device='cuda:0'), covar=tensor([0.0727, 0.1431, 0.2040, 0.1575, 0.0674, 0.1454, 0.0936, 0.0685], + device='cuda:0'), in_proj_covar=tensor([0.0209, 0.0238, 0.0273, 0.0237, 0.0208, 0.0237, 0.0201, 0.0200], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0005, 0.0005], + device='cuda:0') +2023-02-05 20:58:33,546 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0215, 1.4623, 1.3661, 1.2369, 1.6192, 1.4161, 1.3445, 1.5041], + device='cuda:0'), covar=tensor([0.0798, 0.1510, 0.2192, 0.1740, 0.0803, 0.1720, 0.1099, 0.0721], + device='cuda:0'), in_proj_covar=tensor([0.0209, 0.0238, 0.0273, 0.0237, 0.0208, 0.0237, 0.0201, 0.0200], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0005, 0.0005], + device='cuda:0') +2023-02-05 20:58:34,053 INFO [train.py:901] (0/4) Epoch 2, batch 7550, loss[loss=0.3682, simple_loss=0.4054, pruned_loss=0.1655, over 8258.00 frames. ], tot_loss[loss=0.3585, simple_loss=0.3979, pruned_loss=0.1595, over 1613029.15 frames. ], batch size: 24, lr: 2.71e-02, grad_scale: 8.0 +2023-02-05 20:58:46,908 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.6113, 2.0679, 3.4561, 3.0244, 2.7398, 1.9607, 1.5001, 1.7049], + device='cuda:0'), covar=tensor([0.0665, 0.0877, 0.0158, 0.0288, 0.0361, 0.0404, 0.0520, 0.0793], + device='cuda:0'), in_proj_covar=tensor([0.0529, 0.0456, 0.0355, 0.0390, 0.0505, 0.0422, 0.0446, 0.0462], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 20:58:53,883 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.37 vs. limit=5.0 +2023-02-05 20:59:00,904 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=15672.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 20:59:03,264 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-02-05 20:59:05,743 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 +2023-02-05 20:59:08,580 INFO [train.py:901] (0/4) Epoch 2, batch 7600, loss[loss=0.3588, simple_loss=0.4096, pruned_loss=0.154, over 8110.00 frames. ], tot_loss[loss=0.3587, simple_loss=0.398, pruned_loss=0.1597, over 1613868.71 frames. ], batch size: 23, lr: 2.70e-02, grad_scale: 8.0 +2023-02-05 20:59:31,059 INFO [optim.py:369] (0/4) 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 20:59:43,082 INFO [train.py:901] (0/4) Epoch 2, batch 7650, loss[loss=0.383, simple_loss=0.4193, pruned_loss=0.1733, over 8140.00 frames. ], tot_loss[loss=0.359, simple_loss=0.3981, pruned_loss=0.16, over 1615595.19 frames. ], batch size: 22, lr: 2.70e-02, grad_scale: 8.0 +2023-02-05 21:00:19,420 INFO [train.py:901] (0/4) Epoch 2, batch 7700, loss[loss=0.3729, simple_loss=0.4096, pruned_loss=0.1682, over 7813.00 frames. ], tot_loss[loss=0.3563, simple_loss=0.3963, pruned_loss=0.1581, over 1614904.28 frames. ], batch size: 20, lr: 2.69e-02, grad_scale: 8.0 +2023-02-05 21:00:41,047 INFO [optim.py:369] (0/4) 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 21:00:41,248 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0086, 1.2489, 2.2078, 0.9567, 1.8893, 2.4591, 2.2409, 2.0975], + device='cuda:0'), covar=tensor([0.1306, 0.1309, 0.0494, 0.1920, 0.0563, 0.0342, 0.0383, 0.0653], + device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0261, 0.0190, 0.0258, 0.0196, 0.0165, 0.0156, 0.0236], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 21:00:47,292 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6500, 1.4325, 2.9429, 1.1822, 2.0746, 3.2182, 3.0561, 2.6591], + device='cuda:0'), covar=tensor([0.1124, 0.1564, 0.0465, 0.2051, 0.0774, 0.0333, 0.0333, 0.0689], + device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0259, 0.0189, 0.0256, 0.0195, 0.0164, 0.0155, 0.0236], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 21:00:51,193 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 2, batch 7750, loss[loss=0.3544, simple_loss=0.4007, pruned_loss=0.1541, over 8480.00 frames. ], tot_loss[loss=0.3557, simple_loss=0.3963, pruned_loss=0.1576, over 1618998.12 frames. ], batch size: 49, lr: 2.69e-02, grad_scale: 8.0 +2023-02-05 21:01:28,171 INFO [train.py:901] (0/4) Epoch 2, batch 7800, loss[loss=0.3725, simple_loss=0.4137, pruned_loss=0.1657, over 8281.00 frames. ], tot_loss[loss=0.3546, simple_loss=0.3961, pruned_loss=0.1565, over 1619600.78 frames. ], batch size: 23, lr: 2.69e-02, grad_scale: 8.0 +2023-02-05 21:01:40,331 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=15901.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:01:48,105 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=5.39 vs. limit=5.0 +2023-02-05 21:01:50,923 INFO [optim.py:369] (0/4) 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 21:01:59,049 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=15928.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:02:02,909 INFO [train.py:901] (0/4) Epoch 2, batch 7850, loss[loss=0.3284, simple_loss=0.3943, pruned_loss=0.1313, over 8237.00 frames. ], tot_loss[loss=0.3537, simple_loss=0.3953, pruned_loss=0.1561, over 1620196.86 frames. ], batch size: 24, lr: 2.68e-02, grad_scale: 8.0 +2023-02-05 21:02:15,666 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=15953.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:02:36,219 INFO [train.py:901] (0/4) Epoch 2, batch 7900, loss[loss=0.426, simple_loss=0.4572, pruned_loss=0.1974, over 8522.00 frames. ], tot_loss[loss=0.3531, simple_loss=0.395, pruned_loss=0.1556, over 1619086.00 frames. ], batch size: 28, lr: 2.68e-02, grad_scale: 8.0 +2023-02-05 21:02:46,886 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-16000.pt +2023-02-05 21:02:58,246 INFO [optim.py:369] (0/4) 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 21:02:58,743 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 +2023-02-05 21:03:00,385 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16019.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:03:10,216 INFO [train.py:901] (0/4) Epoch 2, batch 7950, loss[loss=0.401, simple_loss=0.4265, pruned_loss=0.1877, over 8468.00 frames. ], tot_loss[loss=0.3533, simple_loss=0.3948, pruned_loss=0.1559, over 1619919.16 frames. ], batch size: 25, lr: 2.68e-02, grad_scale: 8.0 +2023-02-05 21:03:43,334 INFO [train.py:901] (0/4) Epoch 2, batch 8000, loss[loss=0.3664, simple_loss=0.4103, pruned_loss=0.1612, over 8245.00 frames. ], tot_loss[loss=0.3528, simple_loss=0.3942, pruned_loss=0.1557, over 1619693.94 frames. ], batch size: 22, lr: 2.67e-02, grad_scale: 8.0 +2023-02-05 21:03:56,059 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 21:04:16,514 INFO [train.py:901] (0/4) Epoch 2, batch 8050, loss[loss=0.2755, simple_loss=0.3284, pruned_loss=0.1113, over 7565.00 frames. ], tot_loss[loss=0.3513, simple_loss=0.3924, pruned_loss=0.1551, over 1607052.78 frames. ], batch size: 18, lr: 2.67e-02, grad_scale: 8.0 +2023-02-05 21:04:18,269 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=2.05 vs. limit=2.0 +2023-02-05 21:04:39,777 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/epoch-2.pt +2023-02-05 21:04:51,775 WARNING [train.py:1067] (0/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-05 21:04:55,113 INFO [train.py:901] (0/4) Epoch 3, batch 0, loss[loss=0.4537, simple_loss=0.451, pruned_loss=0.2281, over 7700.00 frames. ], tot_loss[loss=0.4537, simple_loss=0.451, pruned_loss=0.2281, over 7700.00 frames. ], batch size: 18, lr: 2.53e-02, grad_scale: 8.0 +2023-02-05 21:04:55,114 INFO [train.py:926] (0/4) Computing validation loss +2023-02-05 21:05:06,956 INFO [train.py:935] (0/4) 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] (0/4) Maximum memory allocated so far is 6524MB +2023-02-05 21:05:07,108 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16167.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:05:23,572 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-05 21:05:42,768 INFO [optim.py:369] (0/4) 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 21:05:42,788 INFO [train.py:901] (0/4) Epoch 3, batch 50, loss[loss=0.3241, simple_loss=0.3541, pruned_loss=0.1471, over 7237.00 frames. ], tot_loss[loss=0.3493, simple_loss=0.3917, pruned_loss=0.1534, over 362656.40 frames. ], batch size: 16, lr: 2.53e-02, grad_scale: 4.0 +2023-02-05 21:05:58,801 WARNING [train.py:1067] (0/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-05 21:06:03,002 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=16245.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:06:18,211 INFO [train.py:901] (0/4) Epoch 3, batch 100, loss[loss=0.3182, simple_loss=0.3761, pruned_loss=0.1302, over 8227.00 frames. ], tot_loss[loss=0.3539, simple_loss=0.3962, pruned_loss=0.1558, over 640706.42 frames. ], batch size: 22, lr: 2.53e-02, grad_scale: 4.0 +2023-02-05 21:06:18,956 WARNING [train.py:1067] (0/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-05 21:06:53,425 INFO [optim.py:369] (0/4) 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 21:06:53,445 INFO [train.py:901] (0/4) Epoch 3, batch 150, loss[loss=0.4067, simple_loss=0.423, pruned_loss=0.1952, over 6573.00 frames. ], tot_loss[loss=0.35, simple_loss=0.3928, pruned_loss=0.1536, over 851914.58 frames. ], batch size: 71, lr: 2.52e-02, grad_scale: 4.0 +2023-02-05 21:06:59,311 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.73 vs. limit=2.0 +2023-02-05 21:07:23,134 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.6260, 1.3665, 5.5045, 2.3636, 4.7608, 4.6793, 5.1159, 5.1143], + device='cuda:0'), covar=tensor([0.0264, 0.3291, 0.0206, 0.1306, 0.0916, 0.0298, 0.0344, 0.0348], + device='cuda:0'), in_proj_covar=tensor([0.0210, 0.0398, 0.0257, 0.0297, 0.0362, 0.0285, 0.0279, 0.0299], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001], + device='cuda:0') +2023-02-05 21:07:23,178 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16360.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:07:24,938 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 3, batch 200, loss[loss=0.3467, simple_loss=0.3715, pruned_loss=0.161, over 7799.00 frames. ], tot_loss[loss=0.3504, simple_loss=0.3927, pruned_loss=0.154, over 1022134.96 frames. ], batch size: 19, lr: 2.52e-02, grad_scale: 4.0 +2023-02-05 21:07:42,144 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16389.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:08:01,475 INFO [optim.py:369] (0/4) 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 21:08:01,496 INFO [train.py:901] (0/4) Epoch 3, batch 250, loss[loss=0.4508, simple_loss=0.4758, pruned_loss=0.2129, over 8458.00 frames. ], tot_loss[loss=0.3521, simple_loss=0.3939, pruned_loss=0.1552, over 1154664.74 frames. ], batch size: 49, lr: 2.51e-02, grad_scale: 4.0 +2023-02-05 21:08:13,913 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-05 21:08:22,564 WARNING [train.py:1067] (0/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-05 21:08:22,623 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=16447.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:08:35,585 INFO [train.py:901] (0/4) Epoch 3, batch 300, loss[loss=0.38, simple_loss=0.4247, pruned_loss=0.1676, over 8762.00 frames. ], tot_loss[loss=0.3515, simple_loss=0.3936, pruned_loss=0.1546, over 1258738.43 frames. ], batch size: 30, lr: 2.51e-02, grad_scale: 4.0 +2023-02-05 21:08:43,567 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16478.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:08:43,709 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 +2023-02-05 21:08:47,101 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.92 vs. limit=2.0 +2023-02-05 21:09:05,161 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=16511.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:09:09,093 INFO [optim.py:369] (0/4) 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 21:09:09,112 INFO [train.py:901] (0/4) Epoch 3, batch 350, loss[loss=0.4263, simple_loss=0.4422, pruned_loss=0.2052, over 6992.00 frames. ], tot_loss[loss=0.3533, simple_loss=0.3955, pruned_loss=0.1556, over 1338637.28 frames. ], batch size: 71, lr: 2.51e-02, grad_scale: 4.0 +2023-02-05 21:09:40,930 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16562.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:09:44,009 INFO [train.py:901] (0/4) Epoch 3, batch 400, loss[loss=0.3643, simple_loss=0.3996, pruned_loss=0.1645, over 7971.00 frames. ], tot_loss[loss=0.3533, simple_loss=0.3956, pruned_loss=0.1555, over 1402353.36 frames. ], batch size: 21, lr: 2.50e-02, grad_scale: 8.0 +2023-02-05 21:10:18,098 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) Epoch 3, batch 450, loss[loss=0.383, simple_loss=0.4205, pruned_loss=0.1727, over 8362.00 frames. ], tot_loss[loss=0.3505, simple_loss=0.3936, pruned_loss=0.1537, over 1447752.95 frames. ], batch size: 24, lr: 2.50e-02, grad_scale: 8.0 +2023-02-05 21:10:24,813 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16626.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:10:35,577 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 3, batch 500, loss[loss=0.3529, simple_loss=0.4078, pruned_loss=0.149, over 8467.00 frames. ], tot_loss[loss=0.349, simple_loss=0.3922, pruned_loss=0.1529, over 1483851.80 frames. ], batch size: 25, lr: 2.50e-02, grad_scale: 8.0 +2023-02-05 21:11:20,368 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-05 21:11:27,935 INFO [optim.py:369] (0/4) 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 21:11:27,955 INFO [train.py:901] (0/4) Epoch 3, batch 550, loss[loss=0.3542, simple_loss=0.4108, pruned_loss=0.1488, over 8471.00 frames. ], tot_loss[loss=0.347, simple_loss=0.3906, pruned_loss=0.1518, over 1513257.59 frames. ], batch size: 25, lr: 2.49e-02, grad_scale: 8.0 +2023-02-05 21:11:38,640 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=16733.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:11:39,433 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16734.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:11:56,568 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16759.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:12:01,692 INFO [train.py:901] (0/4) Epoch 3, batch 600, loss[loss=0.455, simple_loss=0.4561, pruned_loss=0.2269, over 7069.00 frames. ], tot_loss[loss=0.3487, simple_loss=0.392, pruned_loss=0.1527, over 1540310.20 frames. ], batch size: 72, lr: 2.49e-02, grad_scale: 8.0 +2023-02-05 21:12:16,351 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-05 21:12:36,648 INFO [optim.py:369] (0/4) 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 21:12:36,669 INFO [train.py:901] (0/4) Epoch 3, batch 650, loss[loss=0.3727, simple_loss=0.3952, pruned_loss=0.1751, over 8623.00 frames. ], tot_loss[loss=0.3483, simple_loss=0.3916, pruned_loss=0.1526, over 1557241.73 frames. ], batch size: 49, lr: 2.49e-02, grad_scale: 8.0 +2023-02-05 21:12:37,556 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16818.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:12:41,599 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7867, 2.1409, 4.6000, 1.0538, 2.9641, 2.5049, 1.7703, 2.5923], + device='cuda:0'), covar=tensor([0.1146, 0.1443, 0.0410, 0.2235, 0.1062, 0.1457, 0.0963, 0.1777], + device='cuda:0'), in_proj_covar=tensor([0.0422, 0.0395, 0.0458, 0.0482, 0.0527, 0.0460, 0.0423, 0.0526], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:0') +2023-02-05 21:12:54,023 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16843.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:12:57,233 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16848.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:13:10,251 INFO [train.py:901] (0/4) Epoch 3, batch 700, loss[loss=0.3531, simple_loss=0.3846, pruned_loss=0.1608, over 7653.00 frames. ], tot_loss[loss=0.3494, simple_loss=0.392, pruned_loss=0.1534, over 1570103.25 frames. ], batch size: 19, lr: 2.48e-02, grad_scale: 8.0 +2023-02-05 21:13:20,396 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16882.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:13:23,027 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16886.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 21:13:38,462 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16907.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:13:40,030 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.34 vs. limit=5.0 +2023-02-05 21:13:44,831 INFO [optim.py:369] (0/4) 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 21:13:44,858 INFO [train.py:901] (0/4) Epoch 3, batch 750, loss[loss=0.3964, simple_loss=0.4105, pruned_loss=0.1911, over 7927.00 frames. ], tot_loss[loss=0.3479, simple_loss=0.3918, pruned_loss=0.152, over 1584968.59 frames. ], batch size: 20, lr: 2.48e-02, grad_scale: 8.0 +2023-02-05 21:13:49,801 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16924.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:13:59,055 WARNING [train.py:1067] (0/4) 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] (0/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-05 21:14:19,203 INFO [train.py:901] (0/4) Epoch 3, batch 800, loss[loss=0.3039, simple_loss=0.3586, pruned_loss=0.1246, over 7650.00 frames. ], tot_loss[loss=0.3471, simple_loss=0.3909, pruned_loss=0.1516, over 1588217.69 frames. ], batch size: 19, lr: 2.48e-02, grad_scale: 8.0 +2023-02-05 21:14:41,391 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1368, 2.5643, 1.9853, 3.0293, 1.1974, 1.4964, 1.8450, 2.5178], + device='cuda:0'), covar=tensor([0.1058, 0.1168, 0.1610, 0.0483, 0.2410, 0.2534, 0.1971, 0.0929], + device='cuda:0'), in_proj_covar=tensor([0.0310, 0.0324, 0.0313, 0.0230, 0.0299, 0.0325, 0.0347, 0.0304], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0004, 0.0005, 0.0005, 0.0005, 0.0005], + device='cuda:0') +2023-02-05 21:14:53,628 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 3, batch 850, loss[loss=0.364, simple_loss=0.4117, pruned_loss=0.1582, over 8345.00 frames. ], tot_loss[loss=0.3486, simple_loss=0.3917, pruned_loss=0.1527, over 1594767.47 frames. ], batch size: 26, lr: 2.47e-02, grad_scale: 8.0 +2023-02-05 21:15:28,355 INFO [train.py:901] (0/4) Epoch 3, batch 900, loss[loss=0.3168, simple_loss=0.3588, pruned_loss=0.1374, over 7782.00 frames. ], tot_loss[loss=0.3482, simple_loss=0.391, pruned_loss=0.1527, over 1591881.55 frames. ], batch size: 19, lr: 2.47e-02, grad_scale: 8.0 +2023-02-05 21:15:53,772 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 21:16:02,307 INFO [train.py:901] (0/4) Epoch 3, batch 950, loss[loss=0.3601, simple_loss=0.4027, pruned_loss=0.1588, over 8104.00 frames. ], tot_loss[loss=0.3461, simple_loss=0.3897, pruned_loss=0.1513, over 1595890.33 frames. ], batch size: 23, lr: 2.47e-02, grad_scale: 8.0 +2023-02-05 21:16:10,477 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17129.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:16:25,713 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-05 21:16:36,863 INFO [train.py:901] (0/4) Epoch 3, batch 1000, loss[loss=0.3499, simple_loss=0.3921, pruned_loss=0.1539, over 8253.00 frames. ], tot_loss[loss=0.3469, simple_loss=0.3906, pruned_loss=0.1516, over 1603200.33 frames. ], batch size: 22, lr: 2.46e-02, grad_scale: 8.0 +2023-02-05 21:16:57,950 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-05 21:17:03,576 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17207.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:17:10,139 INFO [optim.py:369] (0/4) 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 21:17:10,160 INFO [train.py:901] (0/4) Epoch 3, batch 1050, loss[loss=0.4067, simple_loss=0.4451, pruned_loss=0.1842, over 8038.00 frames. ], tot_loss[loss=0.3496, simple_loss=0.3927, pruned_loss=0.1533, over 1606727.87 frames. ], batch size: 22, lr: 2.46e-02, grad_scale: 8.0 +2023-02-05 21:17:10,172 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-05 21:17:19,690 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=17230.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 21:17:45,189 INFO [train.py:901] (0/4) Epoch 3, batch 1100, loss[loss=0.331, simple_loss=0.3816, pruned_loss=0.1402, over 8470.00 frames. ], tot_loss[loss=0.349, simple_loss=0.3919, pruned_loss=0.153, over 1608258.27 frames. ], batch size: 25, lr: 2.46e-02, grad_scale: 8.0 +2023-02-05 21:17:45,924 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=17268.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:17:51,895 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.1335, 1.2448, 3.2244, 1.0166, 2.7462, 2.7617, 2.9172, 2.8884], + device='cuda:0'), covar=tensor([0.0347, 0.2464, 0.0380, 0.1682, 0.1011, 0.0483, 0.0363, 0.0447], + device='cuda:0'), in_proj_covar=tensor([0.0224, 0.0391, 0.0270, 0.0303, 0.0363, 0.0289, 0.0280, 0.0298], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001], + device='cuda:0') +2023-02-05 21:18:15,933 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8029, 1.4226, 3.0550, 1.2460, 2.2232, 3.4211, 3.2834, 2.9827], + device='cuda:0'), covar=tensor([0.1064, 0.1334, 0.0366, 0.1710, 0.0628, 0.0258, 0.0263, 0.0450], + device='cuda:0'), in_proj_covar=tensor([0.0230, 0.0260, 0.0197, 0.0259, 0.0197, 0.0161, 0.0157, 0.0234], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 21:18:19,090 INFO [optim.py:369] (0/4) 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 21:18:19,110 INFO [train.py:901] (0/4) Epoch 3, batch 1150, loss[loss=0.316, simple_loss=0.3801, pruned_loss=0.126, over 8132.00 frames. ], tot_loss[loss=0.3473, simple_loss=0.3907, pruned_loss=0.1519, over 1608442.53 frames. ], batch size: 22, lr: 2.45e-02, grad_scale: 8.0 +2023-02-05 21:18:22,456 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-05 21:18:24,020 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3373, 2.2060, 3.2087, 0.6977, 3.0966, 2.1502, 1.2395, 1.7025], + device='cuda:0'), covar=tensor([0.0206, 0.0066, 0.0050, 0.0207, 0.0069, 0.0173, 0.0239, 0.0115], + device='cuda:0'), in_proj_covar=tensor([0.0210, 0.0144, 0.0122, 0.0192, 0.0133, 0.0255, 0.0205, 0.0177], + device='cuda:0'), out_proj_covar=tensor([1.1081e-04, 7.4997e-05, 6.3813e-05, 9.7844e-05, 7.0855e-05, 1.4360e-04, + 1.0958e-04, 9.3967e-05], device='cuda:0') +2023-02-05 21:18:38,623 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17345.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 21:18:52,830 INFO [train.py:901] (0/4) Epoch 3, batch 1200, loss[loss=0.3208, simple_loss=0.3766, pruned_loss=0.1325, over 8458.00 frames. ], tot_loss[loss=0.3461, simple_loss=0.3901, pruned_loss=0.151, over 1615496.71 frames. ], batch size: 29, lr: 2.45e-02, grad_scale: 8.0 +2023-02-05 21:18:55,915 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.35 vs. limit=5.0 +2023-02-05 21:19:02,185 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17380.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:19:04,920 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17383.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:19:09,646 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5245, 1.2244, 3.1402, 1.2544, 2.1610, 3.3505, 3.2296, 2.7500], + device='cuda:0'), covar=tensor([0.1306, 0.1749, 0.0421, 0.1986, 0.0792, 0.0362, 0.0344, 0.0704], + device='cuda:0'), in_proj_covar=tensor([0.0225, 0.0258, 0.0194, 0.0253, 0.0196, 0.0161, 0.0159, 0.0233], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 21:19:17,661 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17401.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 21:19:28,365 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 3, batch 1250, loss[loss=0.3069, simple_loss=0.351, pruned_loss=0.1314, over 7657.00 frames. ], tot_loss[loss=0.3465, simple_loss=0.3903, pruned_loss=0.1514, over 1617222.57 frames. ], batch size: 19, lr: 2.45e-02, grad_scale: 8.0 +2023-02-05 21:19:37,226 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.4995, 4.6918, 4.0927, 1.8221, 3.9802, 3.9345, 4.2407, 3.5382], + device='cuda:0'), covar=tensor([0.0793, 0.0502, 0.0931, 0.3568, 0.0483, 0.0572, 0.1066, 0.0609], + device='cuda:0'), in_proj_covar=tensor([0.0345, 0.0228, 0.0275, 0.0356, 0.0249, 0.0199, 0.0252, 0.0188], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], + device='cuda:0') +2023-02-05 21:20:02,612 INFO [train.py:901] (0/4) Epoch 3, batch 1300, loss[loss=0.294, simple_loss=0.3557, pruned_loss=0.1162, over 8246.00 frames. ], tot_loss[loss=0.3476, simple_loss=0.3915, pruned_loss=0.1518, over 1618466.87 frames. ], batch size: 24, lr: 2.44e-02, grad_scale: 8.0 +2023-02-05 21:20:37,546 INFO [train.py:901] (0/4) Epoch 3, batch 1350, loss[loss=0.3196, simple_loss=0.368, pruned_loss=0.1356, over 8082.00 frames. ], tot_loss[loss=0.3491, simple_loss=0.3922, pruned_loss=0.153, over 1619607.84 frames. ], batch size: 21, lr: 2.44e-02, grad_scale: 4.0 +2023-02-05 21:20:37,719 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5689, 1.6124, 3.0017, 1.2883, 2.1269, 3.3297, 2.9770, 2.9658], + device='cuda:0'), covar=tensor([0.1075, 0.1199, 0.0317, 0.1769, 0.0658, 0.0256, 0.0380, 0.0418], + device='cuda:0'), in_proj_covar=tensor([0.0228, 0.0260, 0.0195, 0.0256, 0.0199, 0.0164, 0.0163, 0.0233], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 21:20:38,226 INFO [optim.py:369] (0/4) 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 21:21:00,229 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=17551.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:21:11,044 INFO [train.py:901] (0/4) Epoch 3, batch 1400, loss[loss=0.3351, simple_loss=0.3913, pruned_loss=0.1395, over 8290.00 frames. ], tot_loss[loss=0.3455, simple_loss=0.3897, pruned_loss=0.1507, over 1621414.23 frames. ], batch size: 23, lr: 2.44e-02, grad_scale: 4.0 +2023-02-05 21:21:20,762 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=2.06 vs. limit=2.0 +2023-02-05 21:21:34,781 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17601.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 21:21:46,916 INFO [train.py:901] (0/4) Epoch 3, batch 1450, loss[loss=0.2766, simple_loss=0.3427, pruned_loss=0.1052, over 8079.00 frames. ], tot_loss[loss=0.3432, simple_loss=0.3874, pruned_loss=0.1495, over 1613176.40 frames. ], batch size: 21, lr: 2.43e-02, grad_scale: 4.0 +2023-02-05 21:21:47,591 INFO [optim.py:369] (0/4) 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 21:21:48,927 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-05 21:21:53,220 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17626.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 21:22:02,523 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17639.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:22:19,026 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17664.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:22:20,425 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17666.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:22:20,936 INFO [train.py:901] (0/4) Epoch 3, batch 1500, loss[loss=0.3279, simple_loss=0.3739, pruned_loss=0.1409, over 8098.00 frames. ], tot_loss[loss=0.3424, simple_loss=0.3868, pruned_loss=0.149, over 1611290.37 frames. ], batch size: 23, lr: 2.43e-02, grad_scale: 4.0 +2023-02-05 21:22:56,188 INFO [train.py:901] (0/4) Epoch 3, batch 1550, loss[loss=0.3629, simple_loss=0.4171, pruned_loss=0.1544, over 8494.00 frames. ], tot_loss[loss=0.342, simple_loss=0.3866, pruned_loss=0.1487, over 1613429.75 frames. ], batch size: 28, lr: 2.43e-02, grad_scale: 4.0 +2023-02-05 21:22:56,832 INFO [optim.py:369] (0/4) 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 21:22:56,993 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17718.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:23:01,058 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=17724.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:23:16,063 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=17745.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 21:23:17,954 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17748.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:23:29,331 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=2.05 vs. limit=2.0 +2023-02-05 21:23:30,930 INFO [train.py:901] (0/4) Epoch 3, batch 1600, loss[loss=0.41, simple_loss=0.413, pruned_loss=0.2035, over 6399.00 frames. ], tot_loss[loss=0.3426, simple_loss=0.3868, pruned_loss=0.1492, over 1612864.19 frames. ], batch size: 14, lr: 2.42e-02, grad_scale: 8.0 +2023-02-05 21:23:43,907 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.3133, 2.3309, 1.6209, 1.4471, 1.9807, 1.7617, 2.3830, 1.9943], + device='cuda:0'), covar=tensor([0.0747, 0.1373, 0.2248, 0.1791, 0.0887, 0.1804, 0.1066, 0.0731], + device='cuda:0'), in_proj_covar=tensor([0.0200, 0.0227, 0.0268, 0.0232, 0.0199, 0.0229, 0.0193, 0.0193], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0006, 0.0006, 0.0006, 0.0005, 0.0006, 0.0005, 0.0005], + device='cuda:0') +2023-02-05 21:23:58,114 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4374, 0.9849, 1.0999, 0.8917, 1.2365, 1.0498, 1.1312, 1.0494], + device='cuda:0'), covar=tensor([0.0886, 0.2043, 0.2979, 0.2027, 0.0850, 0.2297, 0.1089, 0.0900], + device='cuda:0'), in_proj_covar=tensor([0.0199, 0.0227, 0.0267, 0.0231, 0.0198, 0.0227, 0.0193, 0.0193], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0006, 0.0006, 0.0006, 0.0005, 0.0006, 0.0005, 0.0005], + device='cuda:0') +2023-02-05 21:24:05,145 INFO [train.py:901] (0/4) Epoch 3, batch 1650, loss[loss=0.3309, simple_loss=0.3914, pruned_loss=0.1351, over 8532.00 frames. ], tot_loss[loss=0.3429, simple_loss=0.388, pruned_loss=0.1489, over 1616559.77 frames. ], batch size: 28, lr: 2.42e-02, grad_scale: 8.0 +2023-02-05 21:24:05,803 INFO [optim.py:369] (0/4) 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 21:24:16,288 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.21 vs. limit=5.0 +2023-02-05 21:24:20,689 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17839.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:24:32,058 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.1792, 1.3170, 1.4575, 1.2691, 0.9007, 1.5110, 0.1046, 0.9395], + device='cuda:0'), covar=tensor([0.1946, 0.1704, 0.0892, 0.1447, 0.2493, 0.1119, 0.3805, 0.1565], + device='cuda:0'), in_proj_covar=tensor([0.0110, 0.0102, 0.0084, 0.0145, 0.0148, 0.0082, 0.0148, 0.0111], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], + device='cuda:0') +2023-02-05 21:24:35,255 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17860.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 21:24:39,698 INFO [train.py:901] (0/4) Epoch 3, batch 1700, loss[loss=0.3653, simple_loss=0.4171, pruned_loss=0.1568, over 8354.00 frames. ], tot_loss[loss=0.3436, simple_loss=0.3886, pruned_loss=0.1493, over 1618028.74 frames. ], batch size: 24, lr: 2.42e-02, grad_scale: 8.0 +2023-02-05 21:25:13,891 INFO [train.py:901] (0/4) Epoch 3, batch 1750, loss[loss=0.35, simple_loss=0.3836, pruned_loss=0.1582, over 7653.00 frames. ], tot_loss[loss=0.3452, simple_loss=0.3899, pruned_loss=0.1503, over 1618172.02 frames. ], batch size: 19, lr: 2.42e-02, grad_scale: 8.0 +2023-02-05 21:25:14,600 INFO [optim.py:369] (0/4) 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 21:25:17,629 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17922.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:25:35,775 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17947.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:25:40,544 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-02-05 21:25:48,886 INFO [train.py:901] (0/4) Epoch 3, batch 1800, loss[loss=0.3535, simple_loss=0.3999, pruned_loss=0.1535, over 8493.00 frames. ], tot_loss[loss=0.3439, simple_loss=0.3892, pruned_loss=0.1493, over 1619395.29 frames. ], batch size: 26, lr: 2.41e-02, grad_scale: 8.0 +2023-02-05 21:25:57,029 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17978.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:26:12,788 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-18000.pt +2023-02-05 21:26:25,063 INFO [train.py:901] (0/4) Epoch 3, batch 1850, loss[loss=0.3104, simple_loss=0.3569, pruned_loss=0.1319, over 7523.00 frames. ], tot_loss[loss=0.3428, simple_loss=0.3878, pruned_loss=0.1489, over 1616642.55 frames. ], batch size: 18, lr: 2.41e-02, grad_scale: 8.0 +2023-02-05 21:26:25,635 INFO [optim.py:369] (0/4) 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 21:26:55,926 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=18062.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:26:59,430 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5731, 1.9014, 1.8669, 1.3948, 1.1607, 1.9262, 0.3532, 1.1461], + device='cuda:0'), covar=tensor([0.3744, 0.1890, 0.1063, 0.1727, 0.3174, 0.0953, 0.3814, 0.1720], + device='cuda:0'), in_proj_covar=tensor([0.0110, 0.0102, 0.0082, 0.0143, 0.0152, 0.0080, 0.0144, 0.0111], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], + device='cuda:0') +2023-02-05 21:26:59,913 INFO [train.py:901] (0/4) Epoch 3, batch 1900, loss[loss=0.3471, simple_loss=0.3796, pruned_loss=0.1573, over 7544.00 frames. ], tot_loss[loss=0.3439, simple_loss=0.3884, pruned_loss=0.1497, over 1620341.25 frames. ], batch size: 18, lr: 2.41e-02, grad_scale: 8.0 +2023-02-05 21:27:17,419 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=18092.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:27:19,599 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18095.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:27:24,176 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-05 21:27:29,725 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.6781, 3.5922, 3.2275, 1.6135, 3.1996, 3.0346, 3.4476, 2.9051], + device='cuda:0'), covar=tensor([0.0860, 0.0647, 0.0982, 0.4327, 0.0657, 0.0803, 0.1085, 0.0723], + device='cuda:0'), in_proj_covar=tensor([0.0337, 0.0232, 0.0279, 0.0364, 0.0246, 0.0208, 0.0261, 0.0188], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001], + device='cuda:0') +2023-02-05 21:27:34,369 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18116.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 21:27:34,803 INFO [train.py:901] (0/4) Epoch 3, batch 1950, loss[loss=0.2911, simple_loss=0.3588, pruned_loss=0.1117, over 8457.00 frames. ], tot_loss[loss=0.3431, simple_loss=0.3877, pruned_loss=0.1492, over 1618936.25 frames. ], batch size: 27, lr: 2.40e-02, grad_scale: 8.0 +2023-02-05 21:27:35,478 INFO [optim.py:369] (0/4) 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 21:27:36,199 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-05 21:27:37,028 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18120.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:27:43,951 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-02-05 21:27:51,192 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18141.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 21:27:55,031 WARNING [train.py:1067] (0/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-05 21:28:09,105 INFO [train.py:901] (0/4) Epoch 3, batch 2000, loss[loss=0.4038, simple_loss=0.4334, pruned_loss=0.1871, over 7820.00 frames. ], tot_loss[loss=0.3458, simple_loss=0.3896, pruned_loss=0.151, over 1617348.09 frames. ], batch size: 20, lr: 2.40e-02, grad_scale: 8.0 +2023-02-05 21:28:16,362 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18177.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:28:27,330 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18192.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:28:27,997 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18193.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:28:38,119 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18207.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:28:41,703 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-05 21:28:44,786 INFO [train.py:901] (0/4) Epoch 3, batch 2050, loss[loss=0.353, simple_loss=0.3928, pruned_loss=0.1566, over 8588.00 frames. ], tot_loss[loss=0.3458, simple_loss=0.3897, pruned_loss=0.1509, over 1618209.88 frames. ], batch size: 39, lr: 2.40e-02, grad_scale: 8.0 +2023-02-05 21:28:46,128 INFO [optim.py:369] (0/4) 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 21:29:18,698 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7523, 2.0717, 1.9036, 2.6378, 1.1084, 1.2986, 1.6505, 2.0438], + device='cuda:0'), covar=tensor([0.1414, 0.1219, 0.1593, 0.0601, 0.2081, 0.2558, 0.2190, 0.1207], + device='cuda:0'), in_proj_covar=tensor([0.0309, 0.0324, 0.0323, 0.0238, 0.0294, 0.0337, 0.0349, 0.0309], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0004, 0.0005, 0.0005, 0.0005, 0.0005], + device='cuda:0') +2023-02-05 21:29:19,884 INFO [train.py:901] (0/4) Epoch 3, batch 2100, loss[loss=0.2864, simple_loss=0.3401, pruned_loss=0.1163, over 8248.00 frames. ], tot_loss[loss=0.3454, simple_loss=0.3893, pruned_loss=0.1508, over 1617057.84 frames. ], batch size: 22, lr: 2.39e-02, grad_scale: 8.0 +2023-02-05 21:29:20,111 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6020, 1.9229, 1.8172, 0.5665, 1.8013, 1.3388, 0.3820, 1.8354], + device='cuda:0'), covar=tensor([0.0114, 0.0063, 0.0073, 0.0149, 0.0091, 0.0261, 0.0211, 0.0062], + device='cuda:0'), in_proj_covar=tensor([0.0213, 0.0149, 0.0126, 0.0190, 0.0145, 0.0265, 0.0209, 0.0177], + device='cuda:0'), out_proj_covar=tensor([1.0889e-04, 7.6253e-05, 6.3287e-05, 9.3655e-05, 7.6118e-05, 1.4493e-04, + 1.0853e-04, 9.1966e-05], device='cuda:0') +2023-02-05 21:29:55,191 INFO [train.py:901] (0/4) Epoch 3, batch 2150, loss[loss=0.352, simple_loss=0.3978, pruned_loss=0.153, over 8678.00 frames. ], tot_loss[loss=0.3454, simple_loss=0.3897, pruned_loss=0.1506, over 1622871.08 frames. ], batch size: 34, lr: 2.39e-02, grad_scale: 8.0 +2023-02-05 21:29:55,884 INFO [optim.py:369] (0/4) 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 21:29:59,195 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=18322.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:30:31,065 INFO [train.py:901] (0/4) Epoch 3, batch 2200, loss[loss=0.3742, simple_loss=0.3894, pruned_loss=0.1795, over 7270.00 frames. ], tot_loss[loss=0.3449, simple_loss=0.3891, pruned_loss=0.1504, over 1615656.11 frames. ], batch size: 16, lr: 2.39e-02, grad_scale: 8.0 +2023-02-05 21:31:06,828 INFO [train.py:901] (0/4) Epoch 3, batch 2250, loss[loss=0.3334, simple_loss=0.3854, pruned_loss=0.1407, over 8106.00 frames. ], tot_loss[loss=0.3443, simple_loss=0.3888, pruned_loss=0.1499, over 1609868.06 frames. ], batch size: 23, lr: 2.38e-02, grad_scale: 8.0 +2023-02-05 21:31:07,494 INFO [optim.py:369] (0/4) 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 21:31:18,175 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18433.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:31:21,446 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18437.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:31:23,565 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2481, 1.5414, 1.4717, 1.2452, 0.8776, 1.5331, 0.1051, 0.9026], + device='cuda:0'), covar=tensor([0.2286, 0.1869, 0.1023, 0.1661, 0.3774, 0.0847, 0.4215, 0.1865], + device='cuda:0'), in_proj_covar=tensor([0.0108, 0.0100, 0.0080, 0.0145, 0.0155, 0.0078, 0.0147, 0.0109], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], + device='cuda:0') +2023-02-05 21:31:33,898 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 +2023-02-05 21:31:36,370 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18458.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:31:38,299 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.7810, 3.8835, 3.3909, 1.7277, 3.3315, 3.1341, 3.4229, 2.8775], + device='cuda:0'), covar=tensor([0.0839, 0.0487, 0.0858, 0.4110, 0.0627, 0.0804, 0.1169, 0.0714], + device='cuda:0'), in_proj_covar=tensor([0.0342, 0.0236, 0.0275, 0.0361, 0.0249, 0.0207, 0.0261, 0.0191], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], + device='cuda:0') +2023-02-05 21:31:39,788 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 3, batch 2300, loss[loss=0.3422, simple_loss=0.3814, pruned_loss=0.1515, over 7801.00 frames. ], tot_loss[loss=0.3435, simple_loss=0.3883, pruned_loss=0.1494, over 1609300.85 frames. ], batch size: 19, lr: 2.38e-02, grad_scale: 8.0 +2023-02-05 21:31:56,792 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18488.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:32:08,041 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-02-05 21:32:10,464 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18508.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:32:17,112 INFO [train.py:901] (0/4) Epoch 3, batch 2350, loss[loss=0.4341, simple_loss=0.4471, pruned_loss=0.2106, over 6869.00 frames. ], tot_loss[loss=0.3406, simple_loss=0.3861, pruned_loss=0.1476, over 1607880.88 frames. ], batch size: 71, lr: 2.38e-02, grad_scale: 8.0 +2023-02-05 21:32:17,761 INFO [optim.py:369] (0/4) 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 21:32:23,358 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18526.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 21:32:30,480 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=18536.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:32:31,154 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=18537.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:32:51,323 INFO [train.py:901] (0/4) Epoch 3, batch 2400, loss[loss=0.3074, simple_loss=0.3528, pruned_loss=0.131, over 7716.00 frames. ], tot_loss[loss=0.3406, simple_loss=0.3863, pruned_loss=0.1474, over 1612651.47 frames. ], batch size: 18, lr: 2.38e-02, grad_scale: 8.0 +2023-02-05 21:33:05,132 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2143, 1.8991, 3.0585, 2.6542, 2.3280, 1.7633, 1.2255, 1.2368], + device='cuda:0'), covar=tensor([0.0844, 0.0929, 0.0173, 0.0334, 0.0398, 0.0514, 0.0793, 0.0795], + device='cuda:0'), in_proj_covar=tensor([0.0575, 0.0494, 0.0398, 0.0446, 0.0544, 0.0463, 0.0482, 0.0488], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 21:33:24,846 INFO [train.py:901] (0/4) Epoch 3, batch 2450, loss[loss=0.3176, simple_loss=0.3822, pruned_loss=0.1265, over 8366.00 frames. ], tot_loss[loss=0.3411, simple_loss=0.3865, pruned_loss=0.1478, over 1611766.70 frames. ], batch size: 24, lr: 2.37e-02, grad_scale: 8.0 +2023-02-05 21:33:25,539 INFO [optim.py:369] (0/4) 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 21:33:33,903 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.3219, 2.4212, 4.3562, 3.5032, 3.0929, 2.5064, 1.6536, 1.9901], + device='cuda:0'), covar=tensor([0.0806, 0.1232, 0.0205, 0.0391, 0.0526, 0.0448, 0.0667, 0.1091], + device='cuda:0'), in_proj_covar=tensor([0.0572, 0.0498, 0.0396, 0.0450, 0.0545, 0.0467, 0.0489, 0.0491], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 21:33:49,161 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18651.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:33:49,834 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18652.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:33:59,424 INFO [train.py:901] (0/4) Epoch 3, batch 2500, loss[loss=0.3575, simple_loss=0.3918, pruned_loss=0.1616, over 7801.00 frames. ], tot_loss[loss=0.3385, simple_loss=0.3843, pruned_loss=0.1463, over 1607269.88 frames. ], batch size: 20, lr: 2.37e-02, grad_scale: 8.0 +2023-02-05 21:34:17,683 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18692.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:34:18,337 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 3, batch 2550, loss[loss=0.3567, simple_loss=0.4027, pruned_loss=0.1554, over 8028.00 frames. ], tot_loss[loss=0.3407, simple_loss=0.386, pruned_loss=0.1477, over 1604578.75 frames. ], batch size: 22, lr: 2.37e-02, grad_scale: 8.0 +2023-02-05 21:34:34,503 INFO [optim.py:369] (0/4) 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 21:34:34,718 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18718.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:35:08,023 INFO [train.py:901] (0/4) Epoch 3, batch 2600, loss[loss=0.3124, simple_loss=0.3555, pruned_loss=0.1347, over 7810.00 frames. ], tot_loss[loss=0.3387, simple_loss=0.3843, pruned_loss=0.1466, over 1604994.76 frames. ], batch size: 20, lr: 2.36e-02, grad_scale: 8.0 +2023-02-05 21:35:37,086 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4748, 1.7355, 1.5641, 1.3463, 1.7113, 1.4381, 1.8387, 1.8975], + device='cuda:0'), covar=tensor([0.0762, 0.1557, 0.1991, 0.1776, 0.0814, 0.1772, 0.1010, 0.0664], + device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0228, 0.0267, 0.0228, 0.0192, 0.0228, 0.0188, 0.0193], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0006, 0.0007, 0.0006, 0.0005, 0.0006, 0.0005, 0.0005], + device='cuda:0') +2023-02-05 21:35:44,462 INFO [train.py:901] (0/4) Epoch 3, batch 2650, loss[loss=0.2872, simple_loss=0.3508, pruned_loss=0.1118, over 8535.00 frames. ], tot_loss[loss=0.3387, simple_loss=0.3838, pruned_loss=0.1468, over 1604664.53 frames. ], batch size: 28, lr: 2.36e-02, grad_scale: 8.0 +2023-02-05 21:35:45,139 INFO [optim.py:369] (0/4) 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 21:36:08,378 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=18852.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:36:19,355 INFO [train.py:901] (0/4) Epoch 3, batch 2700, loss[loss=0.3409, simple_loss=0.4051, pruned_loss=0.1383, over 8035.00 frames. ], tot_loss[loss=0.3377, simple_loss=0.3841, pruned_loss=0.1457, over 1608225.04 frames. ], batch size: 22, lr: 2.36e-02, grad_scale: 8.0 +2023-02-05 21:36:19,548 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9053, 2.2956, 1.8969, 2.7502, 1.6345, 1.5556, 2.0485, 2.4170], + device='cuda:0'), covar=tensor([0.1028, 0.1652, 0.1407, 0.0561, 0.1697, 0.2104, 0.1874, 0.1197], + device='cuda:0'), in_proj_covar=tensor([0.0297, 0.0313, 0.0310, 0.0229, 0.0288, 0.0318, 0.0336, 0.0304], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0004, 0.0005, 0.0005, 0.0005, 0.0005], + device='cuda:0') +2023-02-05 21:36:21,492 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=18870.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 21:36:47,168 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18907.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:36:47,838 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 3, batch 2750, loss[loss=0.351, simple_loss=0.3991, pruned_loss=0.1514, over 8023.00 frames. ], tot_loss[loss=0.3399, simple_loss=0.3859, pruned_loss=0.1469, over 1613040.93 frames. ], batch size: 22, lr: 2.36e-02, grad_scale: 8.0 +2023-02-05 21:36:54,224 INFO [optim.py:369] (0/4) 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 21:36:59,752 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1223, 2.6029, 2.2650, 2.8061, 1.5042, 1.4904, 2.3074, 2.6638], + device='cuda:0'), covar=tensor([0.0919, 0.0979, 0.1178, 0.0531, 0.1481, 0.2003, 0.1390, 0.0730], + device='cuda:0'), in_proj_covar=tensor([0.0293, 0.0311, 0.0310, 0.0230, 0.0290, 0.0315, 0.0336, 0.0301], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0004, 0.0005, 0.0005, 0.0005, 0.0005], + device='cuda:0') +2023-02-05 21:37:00,356 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.5522, 4.6126, 4.1322, 2.0308, 3.9879, 4.1640, 4.2993, 3.6435], + device='cuda:0'), covar=tensor([0.0806, 0.0441, 0.0799, 0.3807, 0.0609, 0.0476, 0.1249, 0.0547], + device='cuda:0'), in_proj_covar=tensor([0.0343, 0.0237, 0.0278, 0.0359, 0.0257, 0.0211, 0.0267, 0.0192], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], + device='cuda:0') +2023-02-05 21:37:04,999 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18932.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:37:05,666 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18933.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:37:15,744 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18948.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:37:28,078 INFO [train.py:901] (0/4) Epoch 3, batch 2800, loss[loss=0.3, simple_loss=0.3642, pruned_loss=0.1179, over 8558.00 frames. ], tot_loss[loss=0.3417, simple_loss=0.3872, pruned_loss=0.1481, over 1614664.82 frames. ], batch size: 31, lr: 2.35e-02, grad_scale: 8.0 +2023-02-05 21:37:28,254 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18967.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:37:41,131 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18985.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 21:37:58,125 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4349, 2.0408, 3.2856, 2.7659, 2.5804, 1.9377, 1.3699, 1.3040], + device='cuda:0'), covar=tensor([0.0876, 0.1031, 0.0187, 0.0381, 0.0437, 0.0492, 0.0608, 0.1043], + device='cuda:0'), in_proj_covar=tensor([0.0576, 0.0502, 0.0399, 0.0450, 0.0555, 0.0464, 0.0487, 0.0494], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 21:38:03,343 INFO [train.py:901] (0/4) Epoch 3, batch 2850, loss[loss=0.368, simple_loss=0.4031, pruned_loss=0.1664, over 8080.00 frames. ], tot_loss[loss=0.3391, simple_loss=0.3855, pruned_loss=0.1464, over 1617109.99 frames. ], batch size: 21, lr: 2.35e-02, grad_scale: 8.0 +2023-02-05 21:38:03,922 INFO [optim.py:369] (0/4) 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 21:38:09,004 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 +2023-02-05 21:38:15,996 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=19036.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:38:37,623 INFO [train.py:901] (0/4) Epoch 3, batch 2900, loss[loss=0.3148, simple_loss=0.3788, pruned_loss=0.1254, over 8287.00 frames. ], tot_loss[loss=0.3395, simple_loss=0.3854, pruned_loss=0.1468, over 1617208.14 frames. ], batch size: 23, lr: 2.35e-02, grad_scale: 8.0 +2023-02-05 21:39:02,485 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-05 21:39:11,757 INFO [train.py:901] (0/4) Epoch 3, batch 2950, loss[loss=0.3596, simple_loss=0.4135, pruned_loss=0.1528, over 8336.00 frames. ], tot_loss[loss=0.3392, simple_loss=0.385, pruned_loss=0.1467, over 1613614.93 frames. ], batch size: 25, lr: 2.34e-02, grad_scale: 8.0 +2023-02-05 21:39:12,413 INFO [optim.py:369] (0/4) 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 21:39:32,497 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19147.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 21:39:35,225 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19151.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:39:46,217 INFO [train.py:901] (0/4) Epoch 3, batch 3000, loss[loss=0.3924, simple_loss=0.4131, pruned_loss=0.1858, over 7694.00 frames. ], tot_loss[loss=0.3399, simple_loss=0.3856, pruned_loss=0.1471, over 1611090.29 frames. ], batch size: 18, lr: 2.34e-02, grad_scale: 8.0 +2023-02-05 21:39:46,218 INFO [train.py:926] (0/4) Computing validation loss +2023-02-05 21:39:58,669 INFO [train.py:935] (0/4) Epoch 3, validation: loss=0.2584, simple_loss=0.3473, pruned_loss=0.08481, over 944034.00 frames. +2023-02-05 21:39:58,670 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6524MB +2023-02-05 21:40:14,606 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7719, 3.8725, 2.1579, 2.5623, 2.4401, 1.7490, 2.2170, 2.5550], + device='cuda:0'), covar=tensor([0.1332, 0.0258, 0.0749, 0.0724, 0.0823, 0.1041, 0.0977, 0.1000], + device='cuda:0'), in_proj_covar=tensor([0.0357, 0.0245, 0.0333, 0.0307, 0.0338, 0.0315, 0.0348, 0.0314], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-05 21:40:33,768 INFO [train.py:901] (0/4) Epoch 3, batch 3050, loss[loss=0.3563, simple_loss=0.4075, pruned_loss=0.1525, over 8192.00 frames. ], tot_loss[loss=0.3417, simple_loss=0.3863, pruned_loss=0.1485, over 1611893.32 frames. ], batch size: 23, lr: 2.34e-02, grad_scale: 8.0 +2023-02-05 21:40:34,455 INFO [optim.py:369] (0/4) 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 21:40:38,077 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=19223.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:40:50,779 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=19241.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 21:40:55,425 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=19248.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:41:07,574 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=19266.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 21:41:08,009 INFO [train.py:901] (0/4) Epoch 3, batch 3100, loss[loss=0.3498, simple_loss=0.3832, pruned_loss=0.1582, over 8074.00 frames. ], tot_loss[loss=0.339, simple_loss=0.3847, pruned_loss=0.1466, over 1614650.83 frames. ], batch size: 21, lr: 2.34e-02, grad_scale: 8.0 +2023-02-05 21:41:26,053 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=19292.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:41:43,833 INFO [train.py:901] (0/4) Epoch 3, batch 3150, loss[loss=0.3331, simple_loss=0.3533, pruned_loss=0.1564, over 7235.00 frames. ], tot_loss[loss=0.3394, simple_loss=0.3851, pruned_loss=0.1468, over 1613701.47 frames. ], batch size: 16, lr: 2.33e-02, grad_scale: 8.0 +2023-02-05 21:41:44,473 INFO [optim.py:369] (0/4) 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 21:42:17,837 INFO [train.py:901] (0/4) Epoch 3, batch 3200, loss[loss=0.2825, simple_loss=0.3428, pruned_loss=0.1111, over 8033.00 frames. ], tot_loss[loss=0.3383, simple_loss=0.3848, pruned_loss=0.1459, over 1617136.90 frames. ], batch size: 22, lr: 2.33e-02, grad_scale: 8.0 +2023-02-05 21:42:45,865 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19407.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:42:45,907 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=19407.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:42:49,228 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19412.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:42:53,526 INFO [train.py:901] (0/4) Epoch 3, batch 3250, loss[loss=0.3304, simple_loss=0.3822, pruned_loss=0.1393, over 8321.00 frames. ], tot_loss[loss=0.3385, simple_loss=0.3852, pruned_loss=0.1459, over 1619313.89 frames. ], batch size: 49, lr: 2.33e-02, grad_scale: 8.0 +2023-02-05 21:42:54,129 INFO [optim.py:369] (0/4) 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 21:43:03,719 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 3, batch 3300, loss[loss=0.3875, simple_loss=0.4134, pruned_loss=0.1807, over 6726.00 frames. ], tot_loss[loss=0.3413, simple_loss=0.3871, pruned_loss=0.1477, over 1617063.58 frames. ], batch size: 71, lr: 2.32e-02, grad_scale: 8.0 +2023-02-05 21:43:43,377 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=19491.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 21:44:00,734 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.96 vs. limit=2.0 +2023-02-05 21:44:01,037 INFO [train.py:901] (0/4) Epoch 3, batch 3350, loss[loss=0.3865, simple_loss=0.4317, pruned_loss=0.1707, over 8639.00 frames. ], tot_loss[loss=0.3406, simple_loss=0.3868, pruned_loss=0.1472, over 1618690.29 frames. ], batch size: 34, lr: 2.32e-02, grad_scale: 16.0 +2023-02-05 21:44:01,702 INFO [optim.py:369] (0/4) 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 21:44:24,038 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=2.03 vs. limit=2.0 +2023-02-05 21:44:35,821 INFO [train.py:901] (0/4) Epoch 3, batch 3400, loss[loss=0.3384, simple_loss=0.39, pruned_loss=0.1434, over 8355.00 frames. ], tot_loss[loss=0.3412, simple_loss=0.3873, pruned_loss=0.1475, over 1615240.43 frames. ], batch size: 26, lr: 2.32e-02, grad_scale: 16.0 +2023-02-05 21:45:02,599 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19606.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 21:45:09,806 INFO [train.py:901] (0/4) Epoch 3, batch 3450, loss[loss=0.3363, simple_loss=0.3915, pruned_loss=0.1405, over 8451.00 frames. ], tot_loss[loss=0.3394, simple_loss=0.3854, pruned_loss=0.1467, over 1613493.45 frames. ], batch size: 25, lr: 2.32e-02, grad_scale: 16.0 +2023-02-05 21:45:10,432 INFO [optim.py:369] (0/4) 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 21:45:11,267 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.6036, 1.0765, 4.6786, 1.9609, 4.0396, 3.8717, 4.1653, 4.1985], + device='cuda:0'), covar=tensor([0.0299, 0.3507, 0.0237, 0.1653, 0.0791, 0.0438, 0.0346, 0.0334], + device='cuda:0'), in_proj_covar=tensor([0.0239, 0.0421, 0.0292, 0.0322, 0.0395, 0.0315, 0.0307, 0.0331], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001], + device='cuda:0') +2023-02-05 21:45:23,378 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19636.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:45:30,355 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-02-05 21:45:42,886 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 3, batch 3500, loss[loss=0.3611, simple_loss=0.413, pruned_loss=0.1546, over 8394.00 frames. ], tot_loss[loss=0.3374, simple_loss=0.3842, pruned_loss=0.1452, over 1613266.22 frames. ], batch size: 49, lr: 2.31e-02, grad_scale: 16.0 +2023-02-05 21:45:58,037 WARNING [train.py:1067] (0/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-05 21:45:59,527 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=19688.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:46:03,405 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5226, 2.5952, 1.7209, 2.2255, 2.1392, 1.2836, 2.0682, 2.1167], + device='cuda:0'), covar=tensor([0.1276, 0.0416, 0.1028, 0.0631, 0.0636, 0.1238, 0.0944, 0.0789], + device='cuda:0'), in_proj_covar=tensor([0.0364, 0.0247, 0.0337, 0.0312, 0.0339, 0.0319, 0.0353, 0.0319], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-05 21:46:19,291 INFO [train.py:901] (0/4) Epoch 3, batch 3550, loss[loss=0.2735, simple_loss=0.3207, pruned_loss=0.1131, over 7690.00 frames. ], tot_loss[loss=0.3346, simple_loss=0.3817, pruned_loss=0.1438, over 1607915.18 frames. ], batch size: 18, lr: 2.31e-02, grad_scale: 16.0 +2023-02-05 21:46:19,968 INFO [optim.py:369] (0/4) 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 21:46:29,856 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 +2023-02-05 21:46:46,351 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=19756.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:46:54,364 INFO [train.py:901] (0/4) Epoch 3, batch 3600, loss[loss=0.3706, simple_loss=0.4176, pruned_loss=0.1619, over 8508.00 frames. ], tot_loss[loss=0.3338, simple_loss=0.3817, pruned_loss=0.143, over 1615124.16 frames. ], batch size: 28, lr: 2.31e-02, grad_scale: 16.0 +2023-02-05 21:47:28,240 INFO [train.py:901] (0/4) Epoch 3, batch 3650, loss[loss=0.312, simple_loss=0.3581, pruned_loss=0.133, over 8505.00 frames. ], tot_loss[loss=0.3354, simple_loss=0.3829, pruned_loss=0.1439, over 1616146.97 frames. ], batch size: 49, lr: 2.30e-02, grad_scale: 16.0 +2023-02-05 21:47:28,896 INFO [optim.py:369] (0/4) 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 21:47:58,730 WARNING [train.py:1067] (0/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-05 21:47:59,604 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=19862.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 21:48:02,764 INFO [train.py:901] (0/4) Epoch 3, batch 3700, loss[loss=0.273, simple_loss=0.3399, pruned_loss=0.1031, over 7797.00 frames. ], tot_loss[loss=0.3354, simple_loss=0.3831, pruned_loss=0.1439, over 1613187.81 frames. ], batch size: 19, lr: 2.30e-02, grad_scale: 16.0 +2023-02-05 21:48:05,662 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19871.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:48:17,450 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=19887.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 21:48:37,846 INFO [train.py:901] (0/4) Epoch 3, batch 3750, loss[loss=0.372, simple_loss=0.4219, pruned_loss=0.1611, over 8453.00 frames. ], tot_loss[loss=0.3357, simple_loss=0.3835, pruned_loss=0.1439, over 1612387.94 frames. ], batch size: 25, lr: 2.30e-02, grad_scale: 16.0 +2023-02-05 21:48:38,362 INFO [optim.py:369] (0/4) 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 21:48:40,669 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.5471, 1.9875, 3.4030, 2.6525, 2.7495, 2.0363, 1.3438, 1.3948], + device='cuda:0'), covar=tensor([0.0979, 0.1248, 0.0242, 0.0523, 0.0542, 0.0572, 0.0718, 0.1249], + device='cuda:0'), in_proj_covar=tensor([0.0601, 0.0513, 0.0421, 0.0466, 0.0574, 0.0482, 0.0501, 0.0504], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 21:48:59,472 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0984, 1.1520, 1.1666, 0.9712, 0.7444, 1.1820, 0.0138, 0.8291], + device='cuda:0'), covar=tensor([0.2766, 0.1849, 0.1128, 0.1880, 0.4790, 0.0900, 0.6254, 0.2177], + device='cuda:0'), in_proj_covar=tensor([0.0111, 0.0095, 0.0081, 0.0149, 0.0169, 0.0077, 0.0154, 0.0108], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], + device='cuda:0') +2023-02-05 21:49:12,035 INFO [train.py:901] (0/4) Epoch 3, batch 3800, loss[loss=0.3861, simple_loss=0.4055, pruned_loss=0.1833, over 7807.00 frames. ], tot_loss[loss=0.3354, simple_loss=0.3833, pruned_loss=0.1437, over 1613163.82 frames. ], batch size: 20, lr: 2.30e-02, grad_scale: 16.0 +2023-02-05 21:49:20,772 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=19980.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:49:34,901 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-20000.pt +2023-02-05 21:49:48,450 INFO [train.py:901] (0/4) Epoch 3, batch 3850, loss[loss=0.3538, simple_loss=0.4015, pruned_loss=0.153, over 8234.00 frames. ], tot_loss[loss=0.3345, simple_loss=0.3822, pruned_loss=0.1434, over 1613536.42 frames. ], batch size: 22, lr: 2.29e-02, grad_scale: 16.0 +2023-02-05 21:49:49,085 INFO [optim.py:369] (0/4) 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 21:49:54,826 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 +2023-02-05 21:50:01,902 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-05 21:50:02,709 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20038.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:50:22,587 INFO [train.py:901] (0/4) Epoch 3, batch 3900, loss[loss=0.3107, simple_loss=0.3617, pruned_loss=0.1299, over 7658.00 frames. ], tot_loss[loss=0.3343, simple_loss=0.3821, pruned_loss=0.1433, over 1615929.68 frames. ], batch size: 19, lr: 2.29e-02, grad_scale: 16.0 +2023-02-05 21:50:41,522 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20095.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:50:50,422 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20107.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:50:56,834 INFO [train.py:901] (0/4) Epoch 3, batch 3950, loss[loss=0.2628, simple_loss=0.3246, pruned_loss=0.1005, over 7419.00 frames. ], tot_loss[loss=0.3355, simple_loss=0.3834, pruned_loss=0.1438, over 1613368.25 frames. ], batch size: 17, lr: 2.29e-02, grad_scale: 16.0 +2023-02-05 21:50:57,395 INFO [optim.py:369] (0/4) 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 21:51:05,035 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20127.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:51:06,967 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20130.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:51:21,454 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 3, batch 4000, loss[loss=0.3075, simple_loss=0.3577, pruned_loss=0.1287, over 8246.00 frames. ], tot_loss[loss=0.3342, simple_loss=0.3822, pruned_loss=0.1431, over 1615080.91 frames. ], batch size: 22, lr: 2.29e-02, grad_scale: 16.0 +2023-02-05 21:51:40,492 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 +2023-02-05 21:51:49,473 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.0533, 2.2501, 3.9586, 3.4064, 2.8217, 2.2756, 1.5105, 2.0097], + device='cuda:0'), covar=tensor([0.0794, 0.1263, 0.0229, 0.0389, 0.0576, 0.0462, 0.0648, 0.1014], + device='cuda:0'), in_proj_covar=tensor([0.0592, 0.0507, 0.0423, 0.0463, 0.0573, 0.0479, 0.0498, 0.0496], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 21:52:05,183 INFO [train.py:901] (0/4) Epoch 3, batch 4050, loss[loss=0.3195, simple_loss=0.3771, pruned_loss=0.1309, over 8665.00 frames. ], tot_loss[loss=0.3328, simple_loss=0.381, pruned_loss=0.1423, over 1615710.55 frames. ], batch size: 39, lr: 2.28e-02, grad_scale: 16.0 +2023-02-05 21:52:05,854 INFO [optim.py:369] (0/4) 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 21:52:09,346 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20222.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:52:40,361 INFO [train.py:901] (0/4) Epoch 3, batch 4100, loss[loss=0.274, simple_loss=0.3359, pruned_loss=0.106, over 7924.00 frames. ], tot_loss[loss=0.3336, simple_loss=0.3819, pruned_loss=0.1426, over 1619653.86 frames. ], batch size: 20, lr: 2.28e-02, grad_scale: 8.0 +2023-02-05 21:52:41,917 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4912, 1.6783, 1.4673, 1.2721, 1.7533, 1.4904, 1.7373, 1.7715], + device='cuda:0'), covar=tensor([0.0806, 0.1570, 0.2396, 0.1878, 0.0889, 0.1978, 0.1011, 0.0763], + device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0224, 0.0262, 0.0224, 0.0189, 0.0227, 0.0185, 0.0187], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0006, 0.0007, 0.0006, 0.0005, 0.0006, 0.0005, 0.0005], + device='cuda:0') +2023-02-05 21:53:14,418 INFO [train.py:901] (0/4) Epoch 3, batch 4150, loss[loss=0.3206, simple_loss=0.3836, pruned_loss=0.1288, over 8528.00 frames. ], tot_loss[loss=0.3328, simple_loss=0.381, pruned_loss=0.1423, over 1620107.08 frames. ], batch size: 26, lr: 2.28e-02, grad_scale: 8.0 +2023-02-05 21:53:15,798 INFO [optim.py:369] (0/4) 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 21:53:20,125 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2871, 2.4332, 1.2020, 1.6970, 1.8562, 1.1070, 1.3985, 2.0348], + device='cuda:0'), covar=tensor([0.1590, 0.0540, 0.1604, 0.0981, 0.1025, 0.1627, 0.1680, 0.1032], + device='cuda:0'), in_proj_covar=tensor([0.0360, 0.0239, 0.0329, 0.0311, 0.0330, 0.0311, 0.0353, 0.0312], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-05 21:53:38,374 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20351.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:53:38,578 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.68 vs. limit=2.0 +2023-02-05 21:53:49,995 INFO [train.py:901] (0/4) Epoch 3, batch 4200, loss[loss=0.3221, simple_loss=0.3742, pruned_loss=0.135, over 7940.00 frames. ], tot_loss[loss=0.3351, simple_loss=0.3823, pruned_loss=0.1439, over 1618662.79 frames. ], batch size: 20, lr: 2.27e-02, grad_scale: 8.0 +2023-02-05 21:53:55,443 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-05 21:53:56,306 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20376.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:54:00,087 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=20382.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:54:14,970 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.7329, 3.7693, 3.3116, 1.3696, 3.2327, 3.1098, 3.4657, 2.8575], + device='cuda:0'), covar=tensor([0.0831, 0.0623, 0.0902, 0.4574, 0.0720, 0.0717, 0.1066, 0.0733], + device='cuda:0'), in_proj_covar=tensor([0.0342, 0.0250, 0.0273, 0.0361, 0.0260, 0.0210, 0.0256, 0.0197], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], + device='cuda:0') +2023-02-05 21:54:16,275 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20406.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:54:16,805 WARNING [train.py:1067] (0/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-05 21:54:24,056 INFO [train.py:901] (0/4) Epoch 3, batch 4250, loss[loss=0.365, simple_loss=0.4088, pruned_loss=0.1606, over 8248.00 frames. ], tot_loss[loss=0.3356, simple_loss=0.3827, pruned_loss=0.1442, over 1618425.15 frames. ], batch size: 24, lr: 2.27e-02, grad_scale: 8.0 +2023-02-05 21:54:24,349 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.78 vs. limit=5.0 +2023-02-05 21:54:25,369 INFO [optim.py:369] (0/4) 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 21:54:37,012 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20436.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:54:47,724 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=20451.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:54:59,257 INFO [train.py:901] (0/4) Epoch 3, batch 4300, loss[loss=0.3123, simple_loss=0.3731, pruned_loss=0.1258, over 7414.00 frames. ], tot_loss[loss=0.3365, simple_loss=0.3839, pruned_loss=0.1446, over 1620443.16 frames. ], batch size: 17, lr: 2.27e-02, grad_scale: 8.0 +2023-02-05 21:55:04,854 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=20474.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:55:20,169 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20497.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:55:33,525 INFO [train.py:901] (0/4) Epoch 3, batch 4350, loss[loss=0.3034, simple_loss=0.3639, pruned_loss=0.1214, over 8086.00 frames. ], tot_loss[loss=0.3356, simple_loss=0.3831, pruned_loss=0.144, over 1620603.00 frames. ], batch size: 21, lr: 2.27e-02, grad_scale: 8.0 +2023-02-05 21:55:34,898 INFO [optim.py:369] (0/4) 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 21:55:46,526 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-05 21:56:06,879 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=20566.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:56:06,985 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20566.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:56:07,509 INFO [train.py:901] (0/4) Epoch 3, batch 4400, loss[loss=0.3903, simple_loss=0.4217, pruned_loss=0.1794, over 8106.00 frames. ], tot_loss[loss=0.3377, simple_loss=0.3843, pruned_loss=0.1456, over 1622953.66 frames. ], batch size: 23, lr: 2.26e-02, grad_scale: 8.0 +2023-02-05 21:56:24,132 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20589.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:56:28,185 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-05 21:56:36,777 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20606.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:56:44,146 INFO [train.py:901] (0/4) Epoch 3, batch 4450, loss[loss=0.3192, simple_loss=0.379, pruned_loss=0.1297, over 8588.00 frames. ], tot_loss[loss=0.334, simple_loss=0.382, pruned_loss=0.143, over 1621409.89 frames. ], batch size: 31, lr: 2.26e-02, grad_scale: 8.0 +2023-02-05 21:56:45,429 INFO [optim.py:369] (0/4) 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 21:57:08,019 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5361, 1.8293, 2.0445, 0.7778, 2.0370, 1.4571, 0.5183, 1.6509], + device='cuda:0'), covar=tensor([0.0136, 0.0094, 0.0054, 0.0164, 0.0102, 0.0258, 0.0223, 0.0082], + device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0163, 0.0136, 0.0212, 0.0161, 0.0282, 0.0221, 0.0189], + device='cuda:0'), out_proj_covar=tensor([1.0942e-04, 7.8161e-05, 6.3258e-05, 9.7395e-05, 7.8224e-05, 1.4447e-04, + 1.0759e-04, 8.9885e-05], device='cuda:0') +2023-02-05 21:57:18,491 INFO [train.py:901] (0/4) Epoch 3, batch 4500, loss[loss=0.2586, simple_loss=0.3108, pruned_loss=0.1032, over 7429.00 frames. ], tot_loss[loss=0.3342, simple_loss=0.3816, pruned_loss=0.1434, over 1617915.47 frames. ], batch size: 17, lr: 2.26e-02, grad_scale: 8.0 +2023-02-05 21:57:20,852 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-05 21:57:27,443 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20681.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:57:46,056 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.2231, 2.1962, 3.9224, 3.6570, 3.3069, 2.5533, 1.6643, 1.8116], + device='cuda:0'), covar=tensor([0.1011, 0.1479, 0.0257, 0.0407, 0.0539, 0.0503, 0.0719, 0.1265], + device='cuda:0'), in_proj_covar=tensor([0.0599, 0.0514, 0.0432, 0.0468, 0.0585, 0.0476, 0.0497, 0.0502], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 21:57:53,171 INFO [train.py:901] (0/4) Epoch 3, batch 4550, loss[loss=0.3024, simple_loss=0.3623, pruned_loss=0.1213, over 7959.00 frames. ], tot_loss[loss=0.3333, simple_loss=0.3807, pruned_loss=0.143, over 1611694.69 frames. ], batch size: 21, lr: 2.26e-02, grad_scale: 8.0 +2023-02-05 21:57:54,484 INFO [optim.py:369] (0/4) 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 21:58:04,580 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 +2023-02-05 21:58:14,820 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=20750.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:58:17,079 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 3, batch 4600, loss[loss=0.3057, simple_loss=0.3528, pruned_loss=0.1293, over 7656.00 frames. ], tot_loss[loss=0.3325, simple_loss=0.3795, pruned_loss=0.1427, over 1608116.46 frames. ], batch size: 19, lr: 2.25e-02, grad_scale: 8.0 +2023-02-05 21:58:34,559 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20778.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:58:34,761 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.35 vs. limit=5.0 +2023-02-05 21:58:35,789 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=20780.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:59:00,498 INFO [train.py:901] (0/4) Epoch 3, batch 4650, loss[loss=0.4144, simple_loss=0.4451, pruned_loss=0.1918, over 8355.00 frames. ], tot_loss[loss=0.3321, simple_loss=0.3796, pruned_loss=0.1423, over 1611159.86 frames. ], batch size: 24, lr: 2.25e-02, grad_scale: 8.0 +2023-02-05 21:59:02,528 INFO [optim.py:369] (0/4) 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 21:59:04,813 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20822.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:59:21,386 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20845.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:59:22,733 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20847.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:59:34,627 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20865.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 21:59:35,790 INFO [train.py:901] (0/4) Epoch 3, batch 4700, loss[loss=0.3037, simple_loss=0.3446, pruned_loss=0.1314, over 7775.00 frames. ], tot_loss[loss=0.3302, simple_loss=0.378, pruned_loss=0.1412, over 1608846.63 frames. ], batch size: 19, lr: 2.25e-02, grad_scale: 8.0 +2023-02-05 21:59:37,902 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20895.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:00:08,985 INFO [train.py:901] (0/4) Epoch 3, batch 4750, loss[loss=0.3189, simple_loss=0.3751, pruned_loss=0.1313, over 8290.00 frames. ], tot_loss[loss=0.3326, simple_loss=0.3798, pruned_loss=0.1426, over 1609423.85 frames. ], batch size: 23, lr: 2.25e-02, grad_scale: 8.0 +2023-02-05 22:00:10,303 INFO [optim.py:369] (0/4) 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 22:00:19,876 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0557, 2.2683, 3.0987, 0.9597, 3.0529, 1.7758, 1.3422, 2.0763], + device='cuda:0'), covar=tensor([0.0153, 0.0079, 0.0062, 0.0216, 0.0103, 0.0220, 0.0228, 0.0096], + device='cuda:0'), in_proj_covar=tensor([0.0227, 0.0163, 0.0133, 0.0207, 0.0155, 0.0273, 0.0222, 0.0191], + device='cuda:0'), out_proj_covar=tensor([1.0768e-04, 7.7221e-05, 6.1022e-05, 9.4978e-05, 7.4801e-05, 1.3865e-04, + 1.0773e-04, 8.9446e-05], device='cuda:0') +2023-02-05 22:00:22,600 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20937.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:00:24,394 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-05 22:00:26,463 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-05 22:00:32,638 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=20950.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:00:41,446 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20962.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:00:44,559 INFO [train.py:901] (0/4) Epoch 3, batch 4800, loss[loss=0.4079, simple_loss=0.4404, pruned_loss=0.1877, over 8428.00 frames. ], tot_loss[loss=0.3327, simple_loss=0.3808, pruned_loss=0.1423, over 1616770.68 frames. ], batch size: 50, lr: 2.24e-02, grad_scale: 8.0 +2023-02-05 22:01:18,122 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-05 22:01:18,802 INFO [train.py:901] (0/4) Epoch 3, batch 4850, loss[loss=0.3291, simple_loss=0.3678, pruned_loss=0.1452, over 7790.00 frames. ], tot_loss[loss=0.3333, simple_loss=0.3804, pruned_loss=0.1431, over 1613110.82 frames. ], batch size: 19, lr: 2.24e-02, grad_scale: 8.0 +2023-02-05 22:01:20,193 INFO [optim.py:369] (0/4) 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 22:01:51,976 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21065.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:01:53,059 INFO [train.py:901] (0/4) Epoch 3, batch 4900, loss[loss=0.4749, simple_loss=0.4821, pruned_loss=0.2338, over 8650.00 frames. ], tot_loss[loss=0.3331, simple_loss=0.3802, pruned_loss=0.143, over 1612699.39 frames. ], batch size: 39, lr: 2.24e-02, grad_scale: 8.0 +2023-02-05 22:01:55,916 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21070.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:02:13,394 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3376, 2.0563, 4.2516, 3.2472, 3.1394, 1.9341, 1.5644, 1.8831], + device='cuda:0'), covar=tensor([0.1669, 0.1852, 0.0267, 0.0561, 0.0706, 0.0909, 0.1059, 0.1549], + device='cuda:0'), in_proj_covar=tensor([0.0603, 0.0520, 0.0434, 0.0487, 0.0589, 0.0488, 0.0501, 0.0513], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 22:02:27,709 INFO [train.py:901] (0/4) Epoch 3, batch 4950, loss[loss=0.3488, simple_loss=0.3895, pruned_loss=0.1541, over 8503.00 frames. ], tot_loss[loss=0.3342, simple_loss=0.3813, pruned_loss=0.1436, over 1617424.20 frames. ], batch size: 28, lr: 2.24e-02, grad_scale: 8.0 +2023-02-05 22:02:29,096 INFO [optim.py:369] (0/4) 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 22:02:30,680 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21121.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:02:41,067 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.78 vs. limit=2.0 +2023-02-05 22:02:41,336 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21136.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:02:48,071 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21146.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:02:51,449 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21151.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:03:01,832 INFO [train.py:901] (0/4) Epoch 3, batch 5000, loss[loss=0.2949, simple_loss=0.3362, pruned_loss=0.1268, over 7648.00 frames. ], tot_loss[loss=0.332, simple_loss=0.3799, pruned_loss=0.142, over 1615717.98 frames. ], batch size: 19, lr: 2.23e-02, grad_scale: 8.0 +2023-02-05 22:03:08,652 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21176.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:03:37,219 INFO [train.py:901] (0/4) Epoch 3, batch 5050, loss[loss=0.3719, simple_loss=0.409, pruned_loss=0.1674, over 8465.00 frames. ], tot_loss[loss=0.3303, simple_loss=0.3783, pruned_loss=0.1412, over 1614223.11 frames. ], batch size: 25, lr: 2.23e-02, grad_scale: 8.0 +2023-02-05 22:03:38,541 INFO [optim.py:369] (0/4) 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] (0/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-05 22:04:11,248 INFO [train.py:901] (0/4) Epoch 3, batch 5100, loss[loss=0.365, simple_loss=0.4055, pruned_loss=0.1623, over 8139.00 frames. ], tot_loss[loss=0.3306, simple_loss=0.3781, pruned_loss=0.1415, over 1610940.88 frames. ], batch size: 22, lr: 2.23e-02, grad_scale: 8.0 +2023-02-05 22:04:46,363 INFO [train.py:901] (0/4) Epoch 3, batch 5150, loss[loss=0.3586, simple_loss=0.4017, pruned_loss=0.1578, over 8023.00 frames. ], tot_loss[loss=0.3298, simple_loss=0.3777, pruned_loss=0.141, over 1610101.22 frames. ], batch size: 22, lr: 2.23e-02, grad_scale: 8.0 +2023-02-05 22:04:47,676 INFO [optim.py:369] (0/4) 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 22:04:49,993 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21321.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:04:52,011 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.5440, 1.1044, 4.5684, 1.5840, 4.0212, 3.6541, 4.0934, 3.9296], + device='cuda:0'), covar=tensor([0.0276, 0.3381, 0.0241, 0.1779, 0.0945, 0.0510, 0.0368, 0.0518], + device='cuda:0'), in_proj_covar=tensor([0.0253, 0.0433, 0.0303, 0.0339, 0.0408, 0.0328, 0.0318, 0.0349], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-05 22:04:56,028 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21330.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:05:06,479 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21346.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:05:06,544 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 3, batch 5200, loss[loss=0.3179, simple_loss=0.3802, pruned_loss=0.1278, over 8191.00 frames. ], tot_loss[loss=0.3305, simple_loss=0.3785, pruned_loss=0.1412, over 1613569.23 frames. ], batch size: 23, lr: 2.22e-02, grad_scale: 8.0 +2023-02-05 22:05:52,823 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 3, batch 5250, loss[loss=0.3554, simple_loss=0.3932, pruned_loss=0.1588, over 7192.00 frames. ], tot_loss[loss=0.3295, simple_loss=0.3776, pruned_loss=0.1407, over 1612597.97 frames. ], batch size: 73, lr: 2.22e-02, grad_scale: 8.0 +2023-02-05 22:05:54,881 WARNING [train.py:1067] (0/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-05 22:05:56,261 INFO [optim.py:369] (0/4) 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 22:06:09,281 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2395, 1.7422, 1.7068, 0.5538, 1.6177, 1.2379, 0.3291, 1.5566], + device='cuda:0'), covar=tensor([0.0124, 0.0063, 0.0066, 0.0136, 0.0089, 0.0250, 0.0196, 0.0068], + device='cuda:0'), in_proj_covar=tensor([0.0223, 0.0159, 0.0134, 0.0207, 0.0161, 0.0279, 0.0220, 0.0196], + device='cuda:0'), out_proj_covar=tensor([1.0445e-04, 7.4113e-05, 6.0828e-05, 9.3271e-05, 7.6606e-05, 1.3987e-04, + 1.0548e-04, 9.1584e-05], device='cuda:0') +2023-02-05 22:06:30,396 INFO [train.py:901] (0/4) Epoch 3, batch 5300, loss[loss=0.305, simple_loss=0.3598, pruned_loss=0.125, over 8476.00 frames. ], tot_loss[loss=0.33, simple_loss=0.3785, pruned_loss=0.1407, over 1611161.23 frames. ], batch size: 27, lr: 2.22e-02, grad_scale: 8.0 +2023-02-05 22:06:39,471 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=21480.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:07:04,794 INFO [train.py:901] (0/4) Epoch 3, batch 5350, loss[loss=0.3836, simple_loss=0.428, pruned_loss=0.1696, over 8360.00 frames. ], tot_loss[loss=0.3297, simple_loss=0.3787, pruned_loss=0.1404, over 1608918.48 frames. ], batch size: 24, lr: 2.22e-02, grad_scale: 8.0 +2023-02-05 22:07:06,076 INFO [optim.py:369] (0/4) 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 22:07:13,627 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21529.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:07:40,163 INFO [train.py:901] (0/4) Epoch 3, batch 5400, loss[loss=0.3547, simple_loss=0.4009, pruned_loss=0.1542, over 8504.00 frames. ], tot_loss[loss=0.3311, simple_loss=0.3796, pruned_loss=0.1413, over 1610468.20 frames. ], batch size: 26, lr: 2.21e-02, grad_scale: 8.0 +2023-02-05 22:07:59,335 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21595.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:08:12,090 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21613.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:08:14,745 INFO [train.py:901] (0/4) Epoch 3, batch 5450, loss[loss=0.3037, simple_loss=0.361, pruned_loss=0.1232, over 8575.00 frames. ], tot_loss[loss=0.3331, simple_loss=0.3814, pruned_loss=0.1424, over 1608564.85 frames. ], batch size: 31, lr: 2.21e-02, grad_scale: 8.0 +2023-02-05 22:08:16,068 INFO [optim.py:369] (0/4) 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 22:08:24,271 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21631.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:08:27,714 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.5677, 4.7096, 3.9644, 1.7249, 3.9527, 4.0251, 4.2625, 3.3511], + device='cuda:0'), covar=tensor([0.0815, 0.0496, 0.1010, 0.4538, 0.0676, 0.0709, 0.1170, 0.0761], + device='cuda:0'), in_proj_covar=tensor([0.0367, 0.0245, 0.0290, 0.0377, 0.0272, 0.0216, 0.0271, 0.0209], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], + device='cuda:0') +2023-02-05 22:08:41,812 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-05 22:08:42,203 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.81 vs. limit=5.0 +2023-02-05 22:08:49,813 INFO [train.py:901] (0/4) Epoch 3, batch 5500, loss[loss=0.3636, simple_loss=0.4025, pruned_loss=0.1624, over 8082.00 frames. ], tot_loss[loss=0.3291, simple_loss=0.3786, pruned_loss=0.1398, over 1609545.50 frames. ], batch size: 21, lr: 2.21e-02, grad_scale: 8.0 +2023-02-05 22:08:55,219 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=21674.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:09:05,879 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=21690.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:09:09,270 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3684, 1.5966, 1.3328, 1.9937, 0.8595, 1.1580, 1.3635, 1.6061], + device='cuda:0'), covar=tensor([0.1262, 0.1210, 0.1567, 0.0563, 0.1602, 0.2124, 0.1361, 0.0967], + device='cuda:0'), in_proj_covar=tensor([0.0306, 0.0307, 0.0303, 0.0229, 0.0287, 0.0313, 0.0329, 0.0296], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0004, 0.0005, 0.0005, 0.0005, 0.0005], + device='cuda:0') +2023-02-05 22:09:23,536 INFO [train.py:901] (0/4) Epoch 3, batch 5550, loss[loss=0.4018, simple_loss=0.4361, pruned_loss=0.1837, over 8483.00 frames. ], tot_loss[loss=0.3303, simple_loss=0.3788, pruned_loss=0.1409, over 1607655.08 frames. ], batch size: 28, lr: 2.21e-02, grad_scale: 8.0 +2023-02-05 22:09:24,903 INFO [optim.py:369] (0/4) 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 22:09:29,325 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 +2023-02-05 22:09:58,420 INFO [train.py:901] (0/4) Epoch 3, batch 5600, loss[loss=0.2959, simple_loss=0.3519, pruned_loss=0.12, over 7924.00 frames. ], tot_loss[loss=0.3307, simple_loss=0.3789, pruned_loss=0.1412, over 1607770.58 frames. ], batch size: 20, lr: 2.20e-02, grad_scale: 8.0 +2023-02-05 22:10:08,483 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21781.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:10:11,824 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21785.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:10:14,437 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21789.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:10:25,392 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21805.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:10:28,816 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21810.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:10:33,420 INFO [train.py:901] (0/4) Epoch 3, batch 5650, loss[loss=0.3192, simple_loss=0.3486, pruned_loss=0.1449, over 7174.00 frames. ], tot_loss[loss=0.3303, simple_loss=0.3785, pruned_loss=0.141, over 1605950.60 frames. ], batch size: 16, lr: 2.20e-02, grad_scale: 8.0 +2023-02-05 22:10:34,793 INFO [optim.py:369] (0/4) 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 22:10:45,255 WARNING [train.py:1067] (0/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-05 22:10:56,856 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21851.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:11:07,103 INFO [train.py:901] (0/4) Epoch 3, batch 5700, loss[loss=0.3408, simple_loss=0.4057, pruned_loss=0.138, over 8248.00 frames. ], tot_loss[loss=0.3286, simple_loss=0.3774, pruned_loss=0.1399, over 1609493.90 frames. ], batch size: 24, lr: 2.20e-02, grad_scale: 8.0 +2023-02-05 22:11:07,924 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21868.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:11:13,376 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 3, batch 5750, loss[loss=0.2747, simple_loss=0.3404, pruned_loss=0.1045, over 8134.00 frames. ], tot_loss[loss=0.3286, simple_loss=0.3777, pruned_loss=0.1398, over 1614714.06 frames. ], batch size: 22, lr: 2.20e-02, grad_scale: 8.0 +2023-02-05 22:11:44,222 INFO [optim.py:369] (0/4) 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 22:11:49,705 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-05 22:12:10,243 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=21957.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:12:16,963 INFO [train.py:901] (0/4) Epoch 3, batch 5800, loss[loss=0.4449, simple_loss=0.4483, pruned_loss=0.2207, over 8603.00 frames. ], tot_loss[loss=0.3302, simple_loss=0.3787, pruned_loss=0.1409, over 1614662.97 frames. ], batch size: 34, lr: 2.19e-02, grad_scale: 8.0 +2023-02-05 22:12:22,428 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=21975.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:12:31,002 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21988.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:12:39,154 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-22000.pt +2023-02-05 22:12:52,159 INFO [train.py:901] (0/4) Epoch 3, batch 5850, loss[loss=0.3503, simple_loss=0.3905, pruned_loss=0.1551, over 8361.00 frames. ], tot_loss[loss=0.3289, simple_loss=0.3775, pruned_loss=0.1401, over 1612924.94 frames. ], batch size: 24, lr: 2.19e-02, grad_scale: 8.0 +2023-02-05 22:12:53,404 INFO [optim.py:369] (0/4) 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 22:13:01,004 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4570, 1.8548, 2.9147, 1.1450, 2.2163, 1.7806, 1.6235, 1.7716], + device='cuda:0'), covar=tensor([0.1237, 0.1421, 0.0509, 0.2402, 0.0962, 0.1806, 0.1110, 0.1505], + device='cuda:0'), in_proj_covar=tensor([0.0443, 0.0417, 0.0487, 0.0505, 0.0548, 0.0485, 0.0430, 0.0557], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:0') +2023-02-05 22:13:02,468 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.86 vs. limit=5.0 +2023-02-05 22:13:11,805 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22045.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:13:22,187 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22061.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:13:25,934 INFO [train.py:901] (0/4) Epoch 3, batch 5900, loss[loss=0.3156, simple_loss=0.3708, pruned_loss=0.1302, over 8078.00 frames. ], tot_loss[loss=0.3282, simple_loss=0.3766, pruned_loss=0.1399, over 1612308.15 frames. ], batch size: 21, lr: 2.19e-02, grad_scale: 8.0 +2023-02-05 22:13:28,807 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22070.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:13:30,185 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22072.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:13:30,260 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7814, 1.6120, 2.4373, 2.0519, 2.0908, 1.6348, 1.2937, 0.8434], + device='cuda:0'), covar=tensor([0.1026, 0.0968, 0.0262, 0.0419, 0.0421, 0.0580, 0.0690, 0.1004], + device='cuda:0'), in_proj_covar=tensor([0.0618, 0.0533, 0.0446, 0.0489, 0.0601, 0.0502, 0.0511, 0.0517], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 22:13:39,564 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22090.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:14:00,337 INFO [train.py:901] (0/4) Epoch 3, batch 5950, loss[loss=0.3717, simple_loss=0.4024, pruned_loss=0.1704, over 8586.00 frames. ], tot_loss[loss=0.3313, simple_loss=0.3787, pruned_loss=0.1419, over 1612177.27 frames. ], batch size: 34, lr: 2.19e-02, grad_scale: 8.0 +2023-02-05 22:14:02,407 INFO [optim.py:369] (0/4) 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 22:14:07,162 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=22125.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:14:22,762 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=5.06 vs. limit=5.0 +2023-02-05 22:14:23,935 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22148.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:14:36,950 INFO [train.py:901] (0/4) Epoch 3, batch 6000, loss[loss=0.3477, simple_loss=0.3963, pruned_loss=0.1496, over 8651.00 frames. ], tot_loss[loss=0.3291, simple_loss=0.3776, pruned_loss=0.1403, over 1616010.43 frames. ], batch size: 34, lr: 2.19e-02, grad_scale: 8.0 +2023-02-05 22:14:36,951 INFO [train.py:926] (0/4) Computing validation loss +2023-02-05 22:14:49,939 INFO [train.py:935] (0/4) Epoch 3, validation: loss=0.2472, simple_loss=0.3383, pruned_loss=0.07805, over 944034.00 frames. +2023-02-05 22:14:49,940 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6689MB +2023-02-05 22:15:08,352 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22194.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:15:21,649 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=22212.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:15:25,121 INFO [train.py:901] (0/4) Epoch 3, batch 6050, loss[loss=0.3312, simple_loss=0.3671, pruned_loss=0.1476, over 7438.00 frames. ], tot_loss[loss=0.3276, simple_loss=0.3765, pruned_loss=0.1393, over 1613516.11 frames. ], batch size: 17, lr: 2.18e-02, grad_scale: 8.0 +2023-02-05 22:15:26,479 INFO [optim.py:369] (0/4) 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] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22233.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:15:40,912 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22240.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:15:51,054 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22255.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:15:59,776 INFO [train.py:901] (0/4) Epoch 3, batch 6100, loss[loss=0.3145, simple_loss=0.3781, pruned_loss=0.1255, over 8366.00 frames. ], tot_loss[loss=0.3302, simple_loss=0.3787, pruned_loss=0.1408, over 1613262.52 frames. ], batch size: 24, lr: 2.18e-02, grad_scale: 16.0 +2023-02-05 22:16:06,254 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.63 vs. limit=5.0 +2023-02-05 22:16:18,457 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-05 22:16:35,129 INFO [train.py:901] (0/4) Epoch 3, batch 6150, loss[loss=0.3369, simple_loss=0.3843, pruned_loss=0.1447, over 8605.00 frames. ], tot_loss[loss=0.3313, simple_loss=0.3796, pruned_loss=0.1415, over 1614343.25 frames. ], batch size: 31, lr: 2.18e-02, grad_scale: 16.0 +2023-02-05 22:16:36,460 INFO [optim.py:369] (0/4) 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 22:16:41,830 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22327.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:16:42,024 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 +2023-02-05 22:16:42,471 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22328.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:16:45,071 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=22332.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:16:54,506 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22346.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:16:59,256 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22353.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:17:08,711 INFO [train.py:901] (0/4) Epoch 3, batch 6200, loss[loss=0.2861, simple_loss=0.3463, pruned_loss=0.113, over 7818.00 frames. ], tot_loss[loss=0.3291, simple_loss=0.3786, pruned_loss=0.1398, over 1619605.70 frames. ], batch size: 20, lr: 2.18e-02, grad_scale: 16.0 +2023-02-05 22:17:11,711 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22371.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:17:32,833 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.53 vs. limit=5.0 +2023-02-05 22:17:34,503 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22403.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:17:44,418 INFO [train.py:901] (0/4) Epoch 3, batch 6250, loss[loss=0.3894, simple_loss=0.4194, pruned_loss=0.1797, over 8692.00 frames. ], tot_loss[loss=0.3292, simple_loss=0.3785, pruned_loss=0.1399, over 1617287.10 frames. ], batch size: 39, lr: 2.17e-02, grad_scale: 16.0 +2023-02-05 22:17:45,750 INFO [optim.py:369] (0/4) 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 22:18:05,739 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22447.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:18:19,135 INFO [train.py:901] (0/4) Epoch 3, batch 6300, loss[loss=0.2676, simple_loss=0.3239, pruned_loss=0.1057, over 7533.00 frames. ], tot_loss[loss=0.3267, simple_loss=0.3763, pruned_loss=0.1386, over 1615797.14 frames. ], batch size: 18, lr: 2.17e-02, grad_scale: 16.0 +2023-02-05 22:18:35,761 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=22492.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:18:39,286 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22496.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:18:54,107 INFO [train.py:901] (0/4) Epoch 3, batch 6350, loss[loss=0.3102, simple_loss=0.3462, pruned_loss=0.1371, over 7719.00 frames. ], tot_loss[loss=0.3241, simple_loss=0.374, pruned_loss=0.1371, over 1612309.80 frames. ], batch size: 18, lr: 2.17e-02, grad_scale: 16.0 +2023-02-05 22:18:55,440 INFO [optim.py:369] (0/4) 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 22:18:57,026 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22521.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:19:08,836 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 3, batch 6400, loss[loss=0.3211, simple_loss=0.3773, pruned_loss=0.1325, over 8032.00 frames. ], tot_loss[loss=0.3235, simple_loss=0.3737, pruned_loss=0.1366, over 1613056.39 frames. ], batch size: 22, lr: 2.17e-02, grad_scale: 16.0 +2023-02-05 22:19:35,377 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=22577.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:19:39,427 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22583.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:19:46,067 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6890, 1.5130, 2.7421, 1.2271, 2.0669, 2.9387, 2.7548, 2.5396], + device='cuda:0'), covar=tensor([0.1112, 0.1338, 0.0453, 0.1958, 0.0647, 0.0328, 0.0404, 0.0651], + device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0263, 0.0200, 0.0260, 0.0209, 0.0179, 0.0176, 0.0249], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 22:19:50,039 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=22599.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:19:55,815 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22607.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:19:56,553 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22608.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:20:03,194 INFO [train.py:901] (0/4) Epoch 3, batch 6450, loss[loss=0.3196, simple_loss=0.3744, pruned_loss=0.1324, over 7810.00 frames. ], tot_loss[loss=0.3226, simple_loss=0.3729, pruned_loss=0.1362, over 1611506.05 frames. ], batch size: 19, lr: 2.16e-02, grad_scale: 16.0 +2023-02-05 22:20:04,473 INFO [optim.py:369] (0/4) 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 22:20:28,474 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22653.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:20:31,157 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22657.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:20:35,791 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3784, 1.9320, 1.9245, 0.8834, 1.9187, 1.4694, 0.3692, 1.7694], + device='cuda:0'), covar=tensor([0.0118, 0.0066, 0.0063, 0.0133, 0.0074, 0.0218, 0.0195, 0.0057], + device='cuda:0'), in_proj_covar=tensor([0.0240, 0.0166, 0.0141, 0.0217, 0.0161, 0.0290, 0.0228, 0.0193], + device='cuda:0'), out_proj_covar=tensor([1.0969e-04, 7.4442e-05, 6.3151e-05, 9.6748e-05, 7.5123e-05, 1.4177e-04, + 1.0452e-04, 8.5482e-05], device='cuda:0') +2023-02-05 22:20:37,604 INFO [train.py:901] (0/4) Epoch 3, batch 6500, loss[loss=0.3155, simple_loss=0.3846, pruned_loss=0.1232, over 8495.00 frames. ], tot_loss[loss=0.3197, simple_loss=0.3705, pruned_loss=0.1345, over 1609619.31 frames. ], batch size: 26, lr: 2.16e-02, grad_scale: 16.0 +2023-02-05 22:20:55,236 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22692.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:21:02,654 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22703.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:21:09,871 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22714.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:21:11,765 INFO [train.py:901] (0/4) Epoch 3, batch 6550, loss[loss=0.3055, simple_loss=0.3577, pruned_loss=0.1267, over 7160.00 frames. ], tot_loss[loss=0.3215, simple_loss=0.3717, pruned_loss=0.1356, over 1606092.13 frames. ], batch size: 16, lr: 2.16e-02, grad_scale: 16.0 +2023-02-05 22:21:13,168 INFO [optim.py:369] (0/4) 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 22:21:19,296 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22728.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:21:28,103 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0793, 1.4926, 1.3887, 1.1790, 1.5698, 1.4011, 1.4133, 1.5612], + device='cuda:0'), covar=tensor([0.0737, 0.1385, 0.1993, 0.1607, 0.0691, 0.1682, 0.0925, 0.0608], + device='cuda:0'), in_proj_covar=tensor([0.0180, 0.0214, 0.0252, 0.0211, 0.0177, 0.0215, 0.0180, 0.0177], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0006, 0.0007, 0.0006, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-05 22:21:28,621 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-05 22:21:32,840 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=22747.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:21:47,206 INFO [train.py:901] (0/4) Epoch 3, batch 6600, loss[loss=0.3766, simple_loss=0.4256, pruned_loss=0.1638, over 8628.00 frames. ], tot_loss[loss=0.3226, simple_loss=0.3733, pruned_loss=0.136, over 1610926.45 frames. ], batch size: 39, lr: 2.16e-02, grad_scale: 8.0 +2023-02-05 22:21:48,605 WARNING [train.py:1067] (0/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-05 22:22:19,046 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22812.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:22:22,356 INFO [train.py:901] (0/4) Epoch 3, batch 6650, loss[loss=0.3483, simple_loss=0.3998, pruned_loss=0.1484, over 8251.00 frames. ], tot_loss[loss=0.3229, simple_loss=0.3737, pruned_loss=0.1361, over 1612192.63 frames. ], batch size: 24, lr: 2.16e-02, grad_scale: 8.0 +2023-02-05 22:22:24,343 INFO [optim.py:369] (0/4) 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 22:22:29,339 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6453, 2.4789, 1.5603, 2.7904, 1.1958, 1.3565, 1.9536, 2.4454], + device='cuda:0'), covar=tensor([0.1576, 0.1099, 0.2681, 0.0621, 0.2094, 0.2416, 0.1696, 0.0973], + device='cuda:0'), in_proj_covar=tensor([0.0306, 0.0304, 0.0305, 0.0230, 0.0281, 0.0313, 0.0317, 0.0299], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0004, 0.0005, 0.0005, 0.0005, 0.0005], + device='cuda:0') +2023-02-05 22:22:40,069 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22843.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:22:53,754 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22862.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:22:54,512 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22863.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:22:57,082 INFO [train.py:901] (0/4) Epoch 3, batch 6700, loss[loss=0.2136, simple_loss=0.277, pruned_loss=0.07506, over 7696.00 frames. ], tot_loss[loss=0.3221, simple_loss=0.3731, pruned_loss=0.1356, over 1611078.95 frames. ], batch size: 18, lr: 2.15e-02, grad_scale: 8.0 +2023-02-05 22:23:09,422 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.08 vs. limit=5.0 +2023-02-05 22:23:12,647 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22888.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:23:26,904 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22909.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:23:32,715 INFO [train.py:901] (0/4) Epoch 3, batch 6750, loss[loss=0.3216, simple_loss=0.3677, pruned_loss=0.1377, over 8361.00 frames. ], tot_loss[loss=0.3224, simple_loss=0.3734, pruned_loss=0.1357, over 1613719.84 frames. ], batch size: 24, lr: 2.15e-02, grad_scale: 8.0 +2023-02-05 22:23:34,754 INFO [optim.py:369] (0/4) 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 22:23:44,512 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22934.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:23:53,616 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22948.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:24:05,893 INFO [train.py:901] (0/4) Epoch 3, batch 6800, loss[loss=0.2587, simple_loss=0.3123, pruned_loss=0.1025, over 7563.00 frames. ], tot_loss[loss=0.3242, simple_loss=0.3742, pruned_loss=0.1371, over 1613471.03 frames. ], batch size: 18, lr: 2.15e-02, grad_scale: 8.0 +2023-02-05 22:24:05,903 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-05 22:24:08,779 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22970.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:24:10,813 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22973.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:24:26,130 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22995.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:24:30,232 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=23001.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:24:41,422 INFO [train.py:901] (0/4) Epoch 3, batch 6850, loss[loss=0.3351, simple_loss=0.3787, pruned_loss=0.1458, over 7195.00 frames. ], tot_loss[loss=0.3238, simple_loss=0.3737, pruned_loss=0.137, over 1611676.81 frames. ], batch size: 16, lr: 2.15e-02, grad_scale: 8.0 +2023-02-05 22:24:43,433 INFO [optim.py:369] (0/4) 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 22:24:54,853 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-05 22:25:13,998 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2555, 1.3244, 2.2657, 0.9882, 2.1648, 2.4181, 2.3100, 2.0918], + device='cuda:0'), covar=tensor([0.1081, 0.1209, 0.0466, 0.2020, 0.0489, 0.0380, 0.0463, 0.0729], + device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0261, 0.0201, 0.0259, 0.0202, 0.0178, 0.0177, 0.0246], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 22:25:15,259 INFO [train.py:901] (0/4) Epoch 3, batch 6900, loss[loss=0.3852, simple_loss=0.4073, pruned_loss=0.1816, over 7223.00 frames. ], tot_loss[loss=0.3247, simple_loss=0.3742, pruned_loss=0.1376, over 1610321.23 frames. ], batch size: 71, lr: 2.14e-02, grad_scale: 8.0 +2023-02-05 22:25:50,208 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23116.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:25:50,692 INFO [train.py:901] (0/4) Epoch 3, batch 6950, loss[loss=0.2973, simple_loss=0.3472, pruned_loss=0.1237, over 8131.00 frames. ], tot_loss[loss=0.3245, simple_loss=0.3742, pruned_loss=0.1375, over 1611235.71 frames. ], batch size: 22, lr: 2.14e-02, grad_scale: 8.0 +2023-02-05 22:25:51,604 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23118.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:25:52,729 INFO [optim.py:369] (0/4) 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 22:25:57,641 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23126.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:26:02,130 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-05 22:26:09,856 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23143.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:26:18,630 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=23156.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:26:26,189 INFO [train.py:901] (0/4) Epoch 3, batch 7000, loss[loss=0.2801, simple_loss=0.3288, pruned_loss=0.1157, over 7409.00 frames. ], tot_loss[loss=0.3252, simple_loss=0.3744, pruned_loss=0.138, over 1608966.12 frames. ], batch size: 17, lr: 2.14e-02, grad_scale: 8.0 +2023-02-05 22:26:39,913 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=23187.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:26:44,867 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4405, 1.7011, 1.5095, 1.2899, 1.5000, 1.5875, 1.7522, 1.6831], + device='cuda:0'), covar=tensor([0.0641, 0.1346, 0.2013, 0.1646, 0.0842, 0.1612, 0.0919, 0.0660], + device='cuda:0'), in_proj_covar=tensor([0.0176, 0.0211, 0.0248, 0.0211, 0.0176, 0.0214, 0.0174, 0.0176], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0006, 0.0007, 0.0006, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-05 22:27:01,198 INFO [train.py:901] (0/4) Epoch 3, batch 7050, loss[loss=0.3613, simple_loss=0.4046, pruned_loss=0.159, over 8027.00 frames. ], tot_loss[loss=0.3271, simple_loss=0.3756, pruned_loss=0.1393, over 1609383.15 frames. ], batch size: 22, lr: 2.14e-02, grad_scale: 8.0 +2023-02-05 22:27:03,862 INFO [optim.py:369] (0/4) 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 22:27:14,101 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23235.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:27:18,797 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.0484, 1.3704, 4.1633, 1.6823, 3.5601, 3.4002, 3.6961, 3.6518], + device='cuda:0'), covar=tensor([0.0309, 0.2983, 0.0299, 0.1666, 0.0965, 0.0500, 0.0369, 0.0395], + device='cuda:0'), in_proj_covar=tensor([0.0263, 0.0429, 0.0321, 0.0340, 0.0405, 0.0334, 0.0316, 0.0356], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001], + device='cuda:0') +2023-02-05 22:27:22,745 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 3, batch 7100, loss[loss=0.356, simple_loss=0.3992, pruned_loss=0.1565, over 7812.00 frames. ], tot_loss[loss=0.3267, simple_loss=0.3755, pruned_loss=0.139, over 1611144.61 frames. ], batch size: 20, lr: 2.14e-02, grad_scale: 8.0 +2023-02-05 22:27:39,094 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23271.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:27:59,141 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23302.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:28:08,841 INFO [train.py:901] (0/4) Epoch 3, batch 7150, loss[loss=0.3096, simple_loss=0.3631, pruned_loss=0.1281, over 8329.00 frames. ], tot_loss[loss=0.3262, simple_loss=0.3757, pruned_loss=0.1383, over 1614416.04 frames. ], batch size: 26, lr: 2.13e-02, grad_scale: 8.0 +2023-02-05 22:28:10,863 INFO [optim.py:369] (0/4) 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 22:28:17,308 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=5.70 vs. limit=5.0 +2023-02-05 22:28:42,756 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9731, 1.7701, 4.1289, 1.7757, 2.4888, 4.6594, 4.3779, 4.1198], + device='cuda:0'), covar=tensor([0.1174, 0.1320, 0.0239, 0.1755, 0.0696, 0.0153, 0.0275, 0.0482], + device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0262, 0.0207, 0.0263, 0.0205, 0.0182, 0.0184, 0.0254], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 22:28:43,292 INFO [train.py:901] (0/4) Epoch 3, batch 7200, loss[loss=0.4062, simple_loss=0.4434, pruned_loss=0.1845, over 8642.00 frames. ], tot_loss[loss=0.328, simple_loss=0.377, pruned_loss=0.1395, over 1610593.85 frames. ], batch size: 34, lr: 2.13e-02, grad_scale: 8.0 +2023-02-05 22:28:47,562 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23372.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:29:04,751 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23397.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:29:17,769 INFO [train.py:901] (0/4) Epoch 3, batch 7250, loss[loss=0.2848, simple_loss=0.3422, pruned_loss=0.1137, over 7802.00 frames. ], tot_loss[loss=0.3275, simple_loss=0.3762, pruned_loss=0.1393, over 1612642.50 frames. ], batch size: 20, lr: 2.13e-02, grad_scale: 4.0 +2023-02-05 22:29:20,312 INFO [optim.py:369] (0/4) 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 22:29:31,973 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.1097, 1.4659, 1.5326, 1.0923, 0.8541, 1.5357, 0.0766, 1.0872], + device='cuda:0'), covar=tensor([0.3674, 0.2366, 0.1538, 0.2764, 0.5916, 0.0918, 0.6310, 0.2371], + device='cuda:0'), in_proj_covar=tensor([0.0120, 0.0115, 0.0086, 0.0159, 0.0183, 0.0080, 0.0150, 0.0118], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 22:29:52,889 INFO [train.py:901] (0/4) Epoch 3, batch 7300, loss[loss=0.3058, simple_loss=0.3609, pruned_loss=0.1253, over 7647.00 frames. ], tot_loss[loss=0.3272, simple_loss=0.3766, pruned_loss=0.1389, over 1614949.50 frames. ], batch size: 19, lr: 2.13e-02, grad_scale: 4.0 +2023-02-05 22:29:55,064 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=23470.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:30:08,356 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.9275, 1.0721, 4.0974, 1.5217, 3.3290, 3.3240, 3.5770, 3.5305], + device='cuda:0'), covar=tensor([0.0510, 0.3765, 0.0370, 0.2059, 0.1355, 0.0581, 0.0492, 0.0644], + device='cuda:0'), in_proj_covar=tensor([0.0268, 0.0439, 0.0321, 0.0349, 0.0417, 0.0346, 0.0320, 0.0368], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-05 22:30:14,526 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.2865, 4.3563, 3.7790, 1.6434, 3.7415, 3.6135, 3.9253, 3.3669], + device='cuda:0'), covar=tensor([0.0780, 0.0516, 0.0839, 0.4435, 0.0634, 0.0744, 0.1193, 0.0645], + device='cuda:0'), in_proj_covar=tensor([0.0363, 0.0249, 0.0276, 0.0365, 0.0269, 0.0223, 0.0267, 0.0208], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], + device='cuda:0') +2023-02-05 22:30:20,012 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.7924, 2.0901, 1.9681, 1.8133, 1.8764, 1.9930, 2.3616, 2.1558], + device='cuda:0'), covar=tensor([0.0529, 0.0941, 0.1452, 0.1284, 0.0660, 0.1199, 0.0660, 0.0473], + device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0208, 0.0243, 0.0210, 0.0171, 0.0211, 0.0174, 0.0174], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0006, 0.0007, 0.0006, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-05 22:30:21,347 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23506.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:30:28,682 INFO [train.py:901] (0/4) Epoch 3, batch 7350, loss[loss=0.2958, simple_loss=0.3528, pruned_loss=0.1194, over 7800.00 frames. ], tot_loss[loss=0.3269, simple_loss=0.3759, pruned_loss=0.139, over 1609445.33 frames. ], batch size: 19, lr: 2.12e-02, grad_scale: 4.0 +2023-02-05 22:30:31,447 INFO [optim.py:369] (0/4) 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 22:30:35,744 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23527.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:30:45,652 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-05 22:30:52,309 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23552.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:30:56,337 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23558.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:31:03,072 INFO [train.py:901] (0/4) Epoch 3, batch 7400, loss[loss=0.3948, simple_loss=0.4388, pruned_loss=0.1754, over 8475.00 frames. ], tot_loss[loss=0.3265, simple_loss=0.3763, pruned_loss=0.1384, over 1610812.34 frames. ], batch size: 29, lr: 2.12e-02, grad_scale: 4.0 +2023-02-05 22:31:03,934 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([0.9834, 1.2110, 1.8014, 0.8654, 1.3695, 1.1525, 1.0605, 1.2062], + device='cuda:0'), covar=tensor([0.0935, 0.1120, 0.0385, 0.1800, 0.0776, 0.1423, 0.0901, 0.1153], + device='cuda:0'), in_proj_covar=tensor([0.0444, 0.0413, 0.0490, 0.0502, 0.0550, 0.0488, 0.0435, 0.0556], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:0') +2023-02-05 22:31:05,768 WARNING [train.py:1067] (0/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-05 22:31:11,874 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=23579.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:31:14,796 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23583.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:31:16,198 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23585.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:31:20,319 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 3, batch 7450, loss[loss=0.4289, simple_loss=0.4597, pruned_loss=0.1991, over 8679.00 frames. ], tot_loss[loss=0.3282, simple_loss=0.3778, pruned_loss=0.1393, over 1615403.75 frames. ], batch size: 34, lr: 2.12e-02, grad_scale: 4.0 +2023-02-05 22:31:41,492 INFO [optim.py:369] (0/4) 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 22:31:44,191 WARNING [train.py:1067] (0/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-05 22:32:03,634 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7352, 1.5944, 3.3662, 1.3889, 2.1618, 3.5989, 3.3202, 3.1055], + device='cuda:0'), covar=tensor([0.1041, 0.1292, 0.0295, 0.1728, 0.0655, 0.0234, 0.0313, 0.0547], + device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0266, 0.0210, 0.0264, 0.0206, 0.0184, 0.0187, 0.0262], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:0') +2023-02-05 22:32:11,862 INFO [train.py:901] (0/4) Epoch 3, batch 7500, loss[loss=0.3516, simple_loss=0.3939, pruned_loss=0.1546, over 7140.00 frames. ], tot_loss[loss=0.3283, simple_loss=0.3777, pruned_loss=0.1395, over 1610814.49 frames. ], batch size: 71, lr: 2.12e-02, grad_scale: 4.0 +2023-02-05 22:32:23,281 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4814, 1.8962, 2.1085, 1.6791, 1.0343, 1.9602, 0.4422, 1.4896], + device='cuda:0'), covar=tensor([0.3600, 0.1861, 0.1289, 0.2284, 0.6641, 0.1115, 0.5974, 0.2020], + device='cuda:0'), in_proj_covar=tensor([0.0122, 0.0115, 0.0084, 0.0159, 0.0187, 0.0082, 0.0147, 0.0119], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 22:32:31,221 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23694.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:32:39,257 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23706.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:32:47,008 INFO [train.py:901] (0/4) Epoch 3, batch 7550, loss[loss=0.2173, simple_loss=0.2897, pruned_loss=0.07251, over 7255.00 frames. ], tot_loss[loss=0.3262, simple_loss=0.376, pruned_loss=0.1382, over 1608829.91 frames. ], batch size: 16, lr: 2.12e-02, grad_scale: 4.0 +2023-02-05 22:32:49,786 INFO [optim.py:369] (0/4) 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 22:33:21,013 INFO [train.py:901] (0/4) Epoch 3, batch 7600, loss[loss=0.3067, simple_loss=0.3799, pruned_loss=0.1168, over 8033.00 frames. ], tot_loss[loss=0.3246, simple_loss=0.3746, pruned_loss=0.1373, over 1608575.87 frames. ], batch size: 22, lr: 2.11e-02, grad_scale: 8.0 +2023-02-05 22:33:55,888 INFO [train.py:901] (0/4) Epoch 3, batch 7650, loss[loss=0.3031, simple_loss=0.3455, pruned_loss=0.1303, over 8082.00 frames. ], tot_loss[loss=0.3244, simple_loss=0.3741, pruned_loss=0.1374, over 1608970.87 frames. ], batch size: 21, lr: 2.11e-02, grad_scale: 8.0 +2023-02-05 22:33:58,400 INFO [optim.py:369] (0/4) 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 22:34:12,405 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 +2023-02-05 22:34:12,868 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23841.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:34:19,390 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=23850.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:34:29,028 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=2.03 vs. limit=2.0 +2023-02-05 22:34:30,201 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23866.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:34:30,651 INFO [train.py:901] (0/4) Epoch 3, batch 7700, loss[loss=0.2973, simple_loss=0.362, pruned_loss=0.1163, over 8367.00 frames. ], tot_loss[loss=0.325, simple_loss=0.3752, pruned_loss=0.1374, over 1614386.11 frames. ], batch size: 24, lr: 2.11e-02, grad_scale: 8.0 +2023-02-05 22:34:50,583 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 +2023-02-05 22:34:50,859 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-05 22:35:04,780 INFO [train.py:901] (0/4) Epoch 3, batch 7750, loss[loss=0.3087, simple_loss=0.3657, pruned_loss=0.1258, over 8324.00 frames. ], tot_loss[loss=0.3259, simple_loss=0.3759, pruned_loss=0.138, over 1615756.04 frames. ], batch size: 25, lr: 2.11e-02, grad_scale: 8.0 +2023-02-05 22:35:08,099 INFO [optim.py:369] (0/4) 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 22:35:27,638 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23950.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:35:37,132 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23962.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:35:39,025 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 3, batch 7800, loss[loss=0.36, simple_loss=0.4064, pruned_loss=0.1567, over 8346.00 frames. ], tot_loss[loss=0.326, simple_loss=0.3761, pruned_loss=0.138, over 1617299.08 frames. ], batch size: 26, lr: 2.11e-02, grad_scale: 8.0 +2023-02-05 22:35:45,535 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23975.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:35:53,464 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23987.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:36:02,019 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-24000.pt +2023-02-05 22:36:14,019 INFO [train.py:901] (0/4) Epoch 3, batch 7850, loss[loss=0.3165, simple_loss=0.3635, pruned_loss=0.1347, over 7809.00 frames. ], tot_loss[loss=0.3268, simple_loss=0.3771, pruned_loss=0.1383, over 1619500.14 frames. ], batch size: 20, lr: 2.10e-02, grad_scale: 8.0 +2023-02-05 22:36:16,543 INFO [optim.py:369] (0/4) 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 22:36:20,032 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.7694, 6.0109, 4.9336, 2.0742, 5.1526, 5.3677, 5.4148, 4.7990], + device='cuda:0'), covar=tensor([0.0714, 0.0300, 0.0883, 0.4578, 0.0508, 0.0444, 0.1082, 0.0561], + device='cuda:0'), in_proj_covar=tensor([0.0359, 0.0246, 0.0286, 0.0369, 0.0273, 0.0230, 0.0270, 0.0214], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], + device='cuda:0') +2023-02-05 22:36:39,258 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=24055.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:36:47,372 INFO [train.py:901] (0/4) Epoch 3, batch 7900, loss[loss=0.3326, simple_loss=0.3799, pruned_loss=0.1427, over 8249.00 frames. ], tot_loss[loss=0.3271, simple_loss=0.3778, pruned_loss=0.1383, over 1622018.17 frames. ], batch size: 22, lr: 2.10e-02, grad_scale: 8.0 +2023-02-05 22:37:20,421 INFO [train.py:901] (0/4) Epoch 3, batch 7950, loss[loss=0.2908, simple_loss=0.3342, pruned_loss=0.1237, over 7698.00 frames. ], tot_loss[loss=0.3235, simple_loss=0.3745, pruned_loss=0.1363, over 1615384.22 frames. ], batch size: 18, lr: 2.10e-02, grad_scale: 8.0 +2023-02-05 22:37:23,174 INFO [optim.py:369] (0/4) 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 22:37:27,563 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.5680, 2.0764, 3.3508, 2.8514, 2.7424, 2.1303, 1.5816, 1.3469], + device='cuda:0'), covar=tensor([0.0936, 0.1195, 0.0252, 0.0502, 0.0539, 0.0580, 0.0695, 0.1253], + device='cuda:0'), in_proj_covar=tensor([0.0632, 0.0548, 0.0464, 0.0517, 0.0628, 0.0509, 0.0523, 0.0536], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 22:37:54,063 INFO [train.py:901] (0/4) Epoch 3, batch 8000, loss[loss=0.3349, simple_loss=0.3809, pruned_loss=0.1445, over 8402.00 frames. ], tot_loss[loss=0.3243, simple_loss=0.3749, pruned_loss=0.1369, over 1615800.44 frames. ], batch size: 48, lr: 2.10e-02, grad_scale: 8.0 +2023-02-05 22:38:25,220 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-02-05 22:38:27,961 INFO [train.py:901] (0/4) Epoch 3, batch 8050, loss[loss=0.2415, simple_loss=0.3127, pruned_loss=0.08514, over 7227.00 frames. ], tot_loss[loss=0.3237, simple_loss=0.3729, pruned_loss=0.1373, over 1588716.38 frames. ], batch size: 16, lr: 2.09e-02, grad_scale: 8.0 +2023-02-05 22:38:30,260 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8917, 1.3145, 1.4046, 1.1363, 1.4161, 1.2950, 1.5673, 1.5470], + device='cuda:0'), covar=tensor([0.0727, 0.1386, 0.2040, 0.1672, 0.0716, 0.1727, 0.0892, 0.0634], + device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0207, 0.0247, 0.0209, 0.0168, 0.0216, 0.0175, 0.0175], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0006, 0.0007, 0.0006, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-05 22:38:30,756 INFO [optim.py:369] (0/4) 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 22:38:31,020 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24221.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:38:46,182 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8088, 1.6619, 2.3114, 1.9633, 2.0403, 1.5609, 1.2367, 1.2161], + device='cuda:0'), covar=tensor([0.0788, 0.0865, 0.0223, 0.0356, 0.0365, 0.0468, 0.0592, 0.0785], + device='cuda:0'), in_proj_covar=tensor([0.0629, 0.0548, 0.0467, 0.0515, 0.0621, 0.0510, 0.0523, 0.0527], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 22:38:48,136 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24246.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:38:51,288 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/epoch-3.pt +2023-02-05 22:39:03,748 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 4, batch 0, loss[loss=0.3673, simple_loss=0.3982, pruned_loss=0.1682, over 8457.00 frames. ], tot_loss[loss=0.3673, simple_loss=0.3982, pruned_loss=0.1682, over 8457.00 frames. ], batch size: 27, lr: 1.96e-02, grad_scale: 8.0 +2023-02-05 22:39:07,720 INFO [train.py:926] (0/4) Computing validation loss +2023-02-05 22:39:18,718 INFO [train.py:935] (0/4) Epoch 4, validation: loss=0.2476, simple_loss=0.3384, pruned_loss=0.07836, over 944034.00 frames. +2023-02-05 22:39:18,719 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6689MB +2023-02-05 22:39:24,332 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.1191, 3.0969, 2.7853, 1.4889, 2.7779, 2.7009, 2.8788, 2.3916], + device='cuda:0'), covar=tensor([0.1262, 0.0799, 0.1173, 0.4282, 0.0897, 0.0936, 0.1489, 0.0996], + device='cuda:0'), in_proj_covar=tensor([0.0364, 0.0248, 0.0291, 0.0374, 0.0277, 0.0234, 0.0271, 0.0216], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:0') +2023-02-05 22:39:34,133 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-05 22:39:52,980 INFO [train.py:901] (0/4) Epoch 4, batch 50, loss[loss=0.3377, simple_loss=0.3911, pruned_loss=0.1421, over 8453.00 frames. ], tot_loss[loss=0.3251, simple_loss=0.3754, pruned_loss=0.1374, over 367592.82 frames. ], batch size: 27, lr: 1.96e-02, grad_scale: 8.0 +2023-02-05 22:40:07,593 INFO [optim.py:369] (0/4) 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 22:40:09,015 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 4, batch 100, loss[loss=0.3389, simple_loss=0.3785, pruned_loss=0.1497, over 8295.00 frames. ], tot_loss[loss=0.324, simple_loss=0.3743, pruned_loss=0.1369, over 644855.47 frames. ], batch size: 23, lr: 1.95e-02, grad_scale: 8.0 +2023-02-05 22:40:31,335 WARNING [train.py:1067] (0/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-05 22:40:43,199 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=2.04 vs. limit=2.0 +2023-02-05 22:40:55,695 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-05 22:41:01,465 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=24399.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:41:02,234 INFO [train.py:901] (0/4) Epoch 4, batch 150, loss[loss=0.3173, simple_loss=0.3686, pruned_loss=0.133, over 8411.00 frames. ], tot_loss[loss=0.3178, simple_loss=0.3706, pruned_loss=0.1325, over 863117.43 frames. ], batch size: 49, lr: 1.95e-02, grad_scale: 8.0 +2023-02-05 22:41:16,207 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.54 vs. limit=5.0 +2023-02-05 22:41:17,163 INFO [optim.py:369] (0/4) 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 22:41:37,211 INFO [train.py:901] (0/4) Epoch 4, batch 200, loss[loss=0.3254, simple_loss=0.3791, pruned_loss=0.1359, over 8252.00 frames. ], tot_loss[loss=0.3182, simple_loss=0.3713, pruned_loss=0.1325, over 1032763.20 frames. ], batch size: 24, lr: 1.95e-02, grad_scale: 8.0 +2023-02-05 22:41:57,240 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 +2023-02-05 22:41:59,067 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5581, 2.0940, 2.2378, 0.9749, 2.1834, 1.3985, 0.6081, 1.8219], + device='cuda:0'), covar=tensor([0.0184, 0.0076, 0.0059, 0.0159, 0.0081, 0.0278, 0.0220, 0.0080], + device='cuda:0'), in_proj_covar=tensor([0.0258, 0.0177, 0.0142, 0.0226, 0.0170, 0.0307, 0.0243, 0.0210], + device='cuda:0'), out_proj_covar=tensor([1.1334e-04, 7.6250e-05, 6.1314e-05, 9.6080e-05, 7.5836e-05, 1.4412e-04, + 1.0749e-04, 8.9656e-05], device='cuda:0') +2023-02-05 22:42:11,042 INFO [train.py:901] (0/4) Epoch 4, batch 250, loss[loss=0.325, simple_loss=0.3714, pruned_loss=0.1393, over 8030.00 frames. ], tot_loss[loss=0.3187, simple_loss=0.3716, pruned_loss=0.1329, over 1163169.78 frames. ], batch size: 22, lr: 1.95e-02, grad_scale: 8.0 +2023-02-05 22:42:12,466 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.5531, 1.0495, 3.9488, 1.6302, 2.8833, 3.2192, 3.6047, 3.5480], + device='cuda:0'), covar=tensor([0.1092, 0.5429, 0.0939, 0.2880, 0.2578, 0.1353, 0.0946, 0.1073], + device='cuda:0'), in_proj_covar=tensor([0.0280, 0.0446, 0.0336, 0.0362, 0.0428, 0.0361, 0.0343, 0.0387], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-05 22:42:13,782 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.2070, 1.1928, 4.3547, 1.6017, 3.7197, 3.6281, 3.9455, 3.8387], + device='cuda:0'), covar=tensor([0.0392, 0.3380, 0.0285, 0.2024, 0.0972, 0.0518, 0.0438, 0.0475], + device='cuda:0'), in_proj_covar=tensor([0.0281, 0.0447, 0.0337, 0.0363, 0.0429, 0.0361, 0.0344, 0.0388], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-05 22:42:20,364 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=24514.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:42:23,593 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-05 22:42:24,846 INFO [optim.py:369] (0/4) 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 22:42:31,609 WARNING [train.py:1067] (0/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-05 22:42:45,996 INFO [train.py:901] (0/4) Epoch 4, batch 300, loss[loss=0.2939, simple_loss=0.3527, pruned_loss=0.1176, over 8497.00 frames. ], tot_loss[loss=0.3211, simple_loss=0.3734, pruned_loss=0.1344, over 1263971.15 frames. ], batch size: 28, lr: 1.95e-02, grad_scale: 8.0 +2023-02-05 22:42:56,998 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=24565.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:43:12,330 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=24587.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:43:21,665 INFO [train.py:901] (0/4) Epoch 4, batch 350, loss[loss=0.3505, simple_loss=0.4044, pruned_loss=0.1483, over 8593.00 frames. ], tot_loss[loss=0.3191, simple_loss=0.3716, pruned_loss=0.1333, over 1341984.07 frames. ], batch size: 39, lr: 1.94e-02, grad_scale: 8.0 +2023-02-05 22:43:31,764 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8254, 2.0964, 2.7630, 1.0461, 2.3917, 1.8490, 1.5559, 1.9397], + device='cuda:0'), covar=tensor([0.0194, 0.0082, 0.0068, 0.0184, 0.0128, 0.0193, 0.0180, 0.0105], + device='cuda:0'), in_proj_covar=tensor([0.0259, 0.0178, 0.0146, 0.0229, 0.0172, 0.0306, 0.0242, 0.0210], + device='cuda:0'), out_proj_covar=tensor([1.1364e-04, 7.6424e-05, 6.2785e-05, 9.6814e-05, 7.6290e-05, 1.4319e-04, + 1.0660e-04, 8.9545e-05], device='cuda:0') +2023-02-05 22:43:34,094 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.64 vs. limit=2.0 +2023-02-05 22:43:35,592 INFO [optim.py:369] (0/4) 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 22:43:52,427 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 +2023-02-05 22:43:56,478 INFO [train.py:901] (0/4) Epoch 4, batch 400, loss[loss=0.3377, simple_loss=0.3797, pruned_loss=0.1479, over 8293.00 frames. ], tot_loss[loss=0.3206, simple_loss=0.3735, pruned_loss=0.1338, over 1407851.75 frames. ], batch size: 23, lr: 1.94e-02, grad_scale: 8.0 +2023-02-05 22:44:04,954 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 +2023-02-05 22:44:13,450 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 +2023-02-05 22:44:30,023 INFO [train.py:901] (0/4) Epoch 4, batch 450, loss[loss=0.2978, simple_loss=0.3653, pruned_loss=0.1152, over 8193.00 frames. ], tot_loss[loss=0.3234, simple_loss=0.376, pruned_loss=0.1354, over 1461069.57 frames. ], batch size: 23, lr: 1.94e-02, grad_scale: 8.0 +2023-02-05 22:44:44,799 INFO [optim.py:369] (0/4) 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 22:45:04,960 INFO [train.py:901] (0/4) Epoch 4, batch 500, loss[loss=0.2939, simple_loss=0.3684, pruned_loss=0.1097, over 8325.00 frames. ], tot_loss[loss=0.3214, simple_loss=0.3744, pruned_loss=0.1342, over 1495997.26 frames. ], batch size: 25, lr: 1.94e-02, grad_scale: 8.0 +2023-02-05 22:45:19,893 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24770.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:45:28,211 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24795.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:45:40,170 INFO [train.py:901] (0/4) Epoch 4, batch 550, loss[loss=0.3532, simple_loss=0.3948, pruned_loss=0.1558, over 8812.00 frames. ], tot_loss[loss=0.3206, simple_loss=0.3736, pruned_loss=0.1338, over 1524840.04 frames. ], batch size: 40, lr: 1.94e-02, grad_scale: 8.0 +2023-02-05 22:45:53,858 INFO [optim.py:369] (0/4) 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 22:46:13,961 INFO [train.py:901] (0/4) Epoch 4, batch 600, loss[loss=0.291, simple_loss=0.3544, pruned_loss=0.1138, over 8099.00 frames. ], tot_loss[loss=0.3188, simple_loss=0.3722, pruned_loss=0.1327, over 1546316.06 frames. ], batch size: 23, lr: 1.93e-02, grad_scale: 8.0 +2023-02-05 22:46:24,931 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=24866.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:46:28,939 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-05 22:46:39,222 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2621, 1.8871, 2.1576, 0.8741, 2.0984, 1.5968, 0.5306, 1.6773], + device='cuda:0'), covar=tensor([0.0150, 0.0068, 0.0053, 0.0152, 0.0096, 0.0211, 0.0212, 0.0074], + device='cuda:0'), in_proj_covar=tensor([0.0252, 0.0175, 0.0140, 0.0219, 0.0165, 0.0297, 0.0236, 0.0207], + device='cuda:0'), 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:0') +2023-02-05 22:46:49,144 INFO [train.py:901] (0/4) Epoch 4, batch 650, loss[loss=0.2878, simple_loss=0.3236, pruned_loss=0.126, over 7687.00 frames. ], tot_loss[loss=0.3196, simple_loss=0.3727, pruned_loss=0.1333, over 1564631.71 frames. ], batch size: 18, lr: 1.93e-02, grad_scale: 8.0 +2023-02-05 22:46:55,178 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=24909.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:47:03,752 INFO [optim.py:369] (0/4) 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 22:47:10,601 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 4, batch 700, loss[loss=0.2965, simple_loss=0.3641, pruned_loss=0.1145, over 8464.00 frames. ], tot_loss[loss=0.3185, simple_loss=0.3719, pruned_loss=0.1325, over 1577429.77 frames. ], batch size: 27, lr: 1.93e-02, grad_scale: 8.0 +2023-02-05 22:47:59,262 INFO [train.py:901] (0/4) Epoch 4, batch 750, loss[loss=0.305, simple_loss=0.3654, pruned_loss=0.1223, over 8293.00 frames. ], tot_loss[loss=0.3179, simple_loss=0.3707, pruned_loss=0.1326, over 1581041.11 frames. ], batch size: 23, lr: 1.93e-02, grad_scale: 8.0 +2023-02-05 22:48:00,335 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.55 vs. limit=5.0 +2023-02-05 22:48:08,255 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25013.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:48:13,344 INFO [optim.py:369] (0/4) 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 22:48:14,041 WARNING [train.py:1067] (0/4) 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] (0/4) 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] (0/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-05 22:48:30,660 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 4, batch 800, loss[loss=0.27, simple_loss=0.3284, pruned_loss=0.1058, over 7714.00 frames. ], tot_loss[loss=0.3177, simple_loss=0.3702, pruned_loss=0.1327, over 1583146.38 frames. ], batch size: 18, lr: 1.93e-02, grad_scale: 8.0 +2023-02-05 22:48:33,437 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.9792, 2.3497, 3.8180, 2.8640, 2.8362, 2.2904, 1.3509, 1.6749], + device='cuda:0'), covar=tensor([0.0948, 0.1243, 0.0260, 0.0670, 0.0759, 0.0590, 0.0801, 0.1370], + device='cuda:0'), in_proj_covar=tensor([0.0655, 0.0563, 0.0488, 0.0538, 0.0663, 0.0525, 0.0539, 0.0550], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 22:48:38,091 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9197, 1.6453, 4.3244, 1.7699, 2.1911, 5.0347, 4.7732, 4.4122], + device='cuda:0'), covar=tensor([0.1322, 0.1477, 0.0290, 0.1929, 0.1055, 0.0221, 0.0295, 0.0511], + device='cuda:0'), in_proj_covar=tensor([0.0230, 0.0254, 0.0204, 0.0261, 0.0206, 0.0182, 0.0185, 0.0259], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:0') +2023-02-05 22:49:06,963 INFO [train.py:901] (0/4) Epoch 4, batch 850, loss[loss=0.3446, simple_loss=0.3653, pruned_loss=0.162, over 8092.00 frames. ], tot_loss[loss=0.3187, simple_loss=0.3705, pruned_loss=0.1334, over 1589347.34 frames. ], batch size: 21, lr: 1.93e-02, grad_scale: 8.0 +2023-02-05 22:49:22,452 INFO [optim.py:369] (0/4) 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 22:49:26,598 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 4, batch 900, loss[loss=0.3222, simple_loss=0.3745, pruned_loss=0.1349, over 8102.00 frames. ], tot_loss[loss=0.3178, simple_loss=0.3705, pruned_loss=0.1325, over 1597738.03 frames. ], batch size: 23, lr: 1.92e-02, grad_scale: 8.0 +2023-02-05 22:50:16,689 INFO [train.py:901] (0/4) Epoch 4, batch 950, loss[loss=0.3155, simple_loss=0.3571, pruned_loss=0.1369, over 7251.00 frames. ], tot_loss[loss=0.3181, simple_loss=0.3715, pruned_loss=0.1323, over 1606324.43 frames. ], batch size: 16, lr: 1.92e-02, grad_scale: 8.0 +2023-02-05 22:50:23,518 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=25210.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:50:30,882 INFO [optim.py:369] (0/4) 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 22:50:40,875 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-05 22:50:46,559 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25242.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:50:51,546 INFO [train.py:901] (0/4) Epoch 4, batch 1000, loss[loss=0.3769, simple_loss=0.4167, pruned_loss=0.1686, over 8290.00 frames. ], tot_loss[loss=0.3192, simple_loss=0.3725, pruned_loss=0.133, over 1613793.48 frames. ], batch size: 23, lr: 1.92e-02, grad_scale: 8.0 +2023-02-05 22:51:12,233 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=25280.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:51:13,323 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-05 22:51:25,837 INFO [train.py:901] (0/4) Epoch 4, batch 1050, loss[loss=0.3176, simple_loss=0.3755, pruned_loss=0.1298, over 8468.00 frames. ], tot_loss[loss=0.3173, simple_loss=0.3709, pruned_loss=0.1318, over 1612656.85 frames. ], batch size: 25, lr: 1.92e-02, grad_scale: 8.0 +2023-02-05 22:51:26,411 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-05 22:51:27,154 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=25302.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:51:28,970 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=25305.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:51:39,506 INFO [optim.py:369] (0/4) 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 22:51:42,387 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=25327.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:51:58,864 INFO [train.py:901] (0/4) Epoch 4, batch 1100, loss[loss=0.3309, simple_loss=0.3796, pruned_loss=0.1411, over 8460.00 frames. ], tot_loss[loss=0.3158, simple_loss=0.3693, pruned_loss=0.1312, over 1611694.60 frames. ], batch size: 27, lr: 1.92e-02, grad_scale: 8.0 +2023-02-05 22:52:04,504 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=25357.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:52:34,677 INFO [train.py:901] (0/4) Epoch 4, batch 1150, loss[loss=0.3249, simple_loss=0.3665, pruned_loss=0.1417, over 7213.00 frames. ], tot_loss[loss=0.3144, simple_loss=0.3681, pruned_loss=0.1304, over 1612523.99 frames. ], batch size: 16, lr: 1.91e-02, grad_scale: 16.0 +2023-02-05 22:52:37,396 WARNING [train.py:1067] (0/4) 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] (0/4) 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 22:53:06,097 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25446.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 22:53:08,498 INFO [train.py:901] (0/4) Epoch 4, batch 1200, loss[loss=0.2994, simple_loss=0.3591, pruned_loss=0.1199, over 8453.00 frames. ], tot_loss[loss=0.3151, simple_loss=0.369, pruned_loss=0.1306, over 1617751.36 frames. ], batch size: 29, lr: 1.91e-02, grad_scale: 16.0 +2023-02-05 22:53:23,223 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25472.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:53:42,023 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=25498.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:53:43,174 INFO [train.py:901] (0/4) Epoch 4, batch 1250, loss[loss=0.2502, simple_loss=0.3012, pruned_loss=0.09958, over 7542.00 frames. ], tot_loss[loss=0.3145, simple_loss=0.368, pruned_loss=0.1305, over 1616771.68 frames. ], batch size: 18, lr: 1.91e-02, grad_scale: 16.0 +2023-02-05 22:53:57,802 INFO [optim.py:369] (0/4) 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 22:53:59,254 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=25523.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:54:18,013 INFO [train.py:901] (0/4) Epoch 4, batch 1300, loss[loss=0.3413, simple_loss=0.3943, pruned_loss=0.1441, over 8247.00 frames. ], tot_loss[loss=0.314, simple_loss=0.3678, pruned_loss=0.1301, over 1614779.22 frames. ], batch size: 24, lr: 1.91e-02, grad_scale: 16.0 +2023-02-05 22:54:32,102 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.65 vs. limit=2.0 +2023-02-05 22:54:39,383 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=25581.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:54:46,394 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-02-05 22:54:53,274 INFO [train.py:901] (0/4) Epoch 4, batch 1350, loss[loss=0.3429, simple_loss=0.393, pruned_loss=0.1463, over 8238.00 frames. ], tot_loss[loss=0.3129, simple_loss=0.3675, pruned_loss=0.1291, over 1619590.90 frames. ], batch size: 24, lr: 1.91e-02, grad_scale: 16.0 +2023-02-05 22:54:57,489 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=25606.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:55:08,863 INFO [optim.py:369] (0/4) 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 22:55:15,332 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3476, 1.5124, 1.2830, 1.9548, 0.7640, 1.0924, 1.1820, 1.5125], + device='cuda:0'), covar=tensor([0.1167, 0.1237, 0.1945, 0.0679, 0.1915, 0.2594, 0.1538, 0.1263], + device='cuda:0'), in_proj_covar=tensor([0.0286, 0.0297, 0.0313, 0.0223, 0.0278, 0.0306, 0.0314, 0.0293], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0004, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-05 22:55:28,882 INFO [train.py:901] (0/4) Epoch 4, batch 1400, loss[loss=0.272, simple_loss=0.3268, pruned_loss=0.1086, over 7970.00 frames. ], tot_loss[loss=0.3146, simple_loss=0.3683, pruned_loss=0.1304, over 1615165.97 frames. ], batch size: 21, lr: 1.91e-02, grad_scale: 8.0 +2023-02-05 22:55:29,831 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5009, 2.0023, 2.0884, 0.5888, 2.1540, 1.3092, 0.5473, 1.7190], + device='cuda:0'), covar=tensor([0.0167, 0.0080, 0.0096, 0.0188, 0.0104, 0.0315, 0.0234, 0.0083], + device='cuda:0'), in_proj_covar=tensor([0.0256, 0.0184, 0.0146, 0.0226, 0.0169, 0.0301, 0.0252, 0.0211], + device='cuda:0'), out_proj_covar=tensor([1.0935e-04, 7.7768e-05, 6.0936e-05, 9.3158e-05, 7.2259e-05, 1.3835e-04, + 1.0928e-04, 8.8565e-05], device='cuda:0') +2023-02-05 22:55:43,751 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 +2023-02-05 22:56:03,149 INFO [train.py:901] (0/4) Epoch 4, batch 1450, loss[loss=0.305, simple_loss=0.3505, pruned_loss=0.1297, over 8090.00 frames. ], tot_loss[loss=0.3129, simple_loss=0.3665, pruned_loss=0.1296, over 1612101.62 frames. ], batch size: 21, lr: 1.90e-02, grad_scale: 8.0 +2023-02-05 22:56:05,833 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-05 22:56:18,895 INFO [optim.py:369] (0/4) 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 22:56:23,169 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=25728.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:56:38,593 INFO [train.py:901] (0/4) Epoch 4, batch 1500, loss[loss=0.2856, simple_loss=0.3415, pruned_loss=0.1149, over 8127.00 frames. ], tot_loss[loss=0.3128, simple_loss=0.3668, pruned_loss=0.1294, over 1611400.37 frames. ], batch size: 22, lr: 1.90e-02, grad_scale: 8.0 +2023-02-05 22:56:40,766 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=25753.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 22:57:02,622 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.2865, 4.2025, 3.7787, 1.5774, 3.8059, 3.6188, 4.0514, 3.3121], + device='cuda:0'), covar=tensor([0.0949, 0.0543, 0.0869, 0.4639, 0.0662, 0.0741, 0.1052, 0.0789], + device='cuda:0'), in_proj_covar=tensor([0.0380, 0.0254, 0.0300, 0.0381, 0.0294, 0.0236, 0.0280, 0.0226], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:0') +2023-02-05 22:57:05,978 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=25790.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 22:57:12,690 INFO [train.py:901] (0/4) Epoch 4, batch 1550, loss[loss=0.2926, simple_loss=0.3539, pruned_loss=0.1157, over 8136.00 frames. ], tot_loss[loss=0.3126, simple_loss=0.3666, pruned_loss=0.1292, over 1614361.90 frames. ], batch size: 22, lr: 1.90e-02, grad_scale: 8.0 +2023-02-05 22:57:27,004 INFO [optim.py:369] (0/4) 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 22:57:46,730 INFO [train.py:901] (0/4) Epoch 4, batch 1600, loss[loss=0.3302, simple_loss=0.3963, pruned_loss=0.1321, over 8569.00 frames. ], tot_loss[loss=0.3143, simple_loss=0.3675, pruned_loss=0.1306, over 1610026.07 frames. ], batch size: 31, lr: 1.90e-02, grad_scale: 8.0 +2023-02-05 22:58:04,744 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25876.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 22:58:21,015 INFO [train.py:901] (0/4) Epoch 4, batch 1650, loss[loss=0.3737, simple_loss=0.4187, pruned_loss=0.1643, over 8433.00 frames. ], tot_loss[loss=0.3159, simple_loss=0.3689, pruned_loss=0.1315, over 1611805.57 frames. ], batch size: 29, lr: 1.90e-02, grad_scale: 8.0 +2023-02-05 22:58:24,577 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25905.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 22:58:35,944 INFO [optim.py:369] (0/4) 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 22:58:56,110 INFO [train.py:901] (0/4) Epoch 4, batch 1700, loss[loss=0.3093, simple_loss=0.3551, pruned_loss=0.1317, over 7792.00 frames. ], tot_loss[loss=0.3183, simple_loss=0.3704, pruned_loss=0.1331, over 1612277.88 frames. ], batch size: 19, lr: 1.90e-02, grad_scale: 8.0 +2023-02-05 22:59:14,216 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.57 vs. limit=5.0 +2023-02-05 22:59:31,297 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-26000.pt +2023-02-05 22:59:32,226 INFO [train.py:901] (0/4) Epoch 4, batch 1750, loss[loss=0.2731, simple_loss=0.3326, pruned_loss=0.1068, over 8248.00 frames. ], tot_loss[loss=0.3167, simple_loss=0.3694, pruned_loss=0.132, over 1616882.45 frames. ], batch size: 24, lr: 1.89e-02, grad_scale: 8.0 +2023-02-05 22:59:47,004 INFO [optim.py:369] (0/4) 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 23:00:06,096 INFO [train.py:901] (0/4) Epoch 4, batch 1800, loss[loss=0.3253, simple_loss=0.3695, pruned_loss=0.1405, over 8132.00 frames. ], tot_loss[loss=0.3156, simple_loss=0.369, pruned_loss=0.1311, over 1618350.92 frames. ], batch size: 22, lr: 1.89e-02, grad_scale: 8.0 +2023-02-05 23:00:30,841 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0520, 2.5365, 1.8587, 2.9989, 1.5039, 1.3295, 1.8502, 2.5553], + device='cuda:0'), covar=tensor([0.1004, 0.1280, 0.1591, 0.0429, 0.1823, 0.2532, 0.1678, 0.0922], + device='cuda:0'), in_proj_covar=tensor([0.0290, 0.0289, 0.0301, 0.0222, 0.0270, 0.0304, 0.0311, 0.0280], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0004, 0.0005, 0.0005, 0.0005, 0.0005], + device='cuda:0') +2023-02-05 23:00:41,281 INFO [train.py:901] (0/4) Epoch 4, batch 1850, loss[loss=0.3001, simple_loss=0.3569, pruned_loss=0.1216, over 8105.00 frames. ], tot_loss[loss=0.3159, simple_loss=0.3693, pruned_loss=0.1312, over 1618219.34 frames. ], batch size: 23, lr: 1.89e-02, grad_scale: 8.0 +2023-02-05 23:00:55,431 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26120.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:00:56,607 INFO [optim.py:369] (0/4) 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 23:01:15,405 INFO [train.py:901] (0/4) Epoch 4, batch 1900, loss[loss=0.3249, simple_loss=0.3867, pruned_loss=0.1315, over 8623.00 frames. ], tot_loss[loss=0.3155, simple_loss=0.3693, pruned_loss=0.1308, over 1621379.56 frames. ], batch size: 34, lr: 1.89e-02, grad_scale: 8.0 +2023-02-05 23:01:16,829 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1415, 2.5932, 1.9609, 3.0690, 1.5528, 1.4913, 1.7932, 2.6959], + device='cuda:0'), covar=tensor([0.1213, 0.1358, 0.1746, 0.0459, 0.1965, 0.2565, 0.2038, 0.0996], + device='cuda:0'), in_proj_covar=tensor([0.0291, 0.0291, 0.0302, 0.0225, 0.0272, 0.0305, 0.0312, 0.0281], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0004, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-05 23:01:22,866 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26161.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 23:01:36,364 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.10 vs. limit=5.0 +2023-02-05 23:01:40,595 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26186.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:01:40,657 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26186.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 23:01:41,086 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-05 23:01:49,747 INFO [train.py:901] (0/4) Epoch 4, batch 1950, loss[loss=0.359, simple_loss=0.389, pruned_loss=0.1646, over 8489.00 frames. ], tot_loss[loss=0.3165, simple_loss=0.3701, pruned_loss=0.1314, over 1622916.16 frames. ], batch size: 49, lr: 1.89e-02, grad_scale: 8.0 +2023-02-05 23:01:52,432 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-05 23:02:04,041 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=26220.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 23:02:05,153 INFO [optim.py:369] (0/4) 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 23:02:10,411 WARNING [train.py:1067] (0/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-05 23:02:24,334 INFO [train.py:901] (0/4) Epoch 4, batch 2000, loss[loss=0.3458, simple_loss=0.3958, pruned_loss=0.1479, over 8194.00 frames. ], tot_loss[loss=0.3181, simple_loss=0.371, pruned_loss=0.1326, over 1621107.20 frames. ], batch size: 23, lr: 1.88e-02, grad_scale: 8.0 +2023-02-05 23:02:26,518 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.42 vs. limit=5.0 +2023-02-05 23:02:34,556 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.6775, 1.2771, 3.8819, 1.3362, 3.3618, 3.2491, 3.4527, 3.3649], + device='cuda:0'), covar=tensor([0.0455, 0.2875, 0.0344, 0.2015, 0.0920, 0.0518, 0.0451, 0.0574], + device='cuda:0'), in_proj_covar=tensor([0.0278, 0.0444, 0.0339, 0.0361, 0.0429, 0.0361, 0.0342, 0.0383], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-05 23:02:36,646 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26268.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:02:59,488 INFO [train.py:901] (0/4) Epoch 4, batch 2050, loss[loss=0.2966, simple_loss=0.3385, pruned_loss=0.1273, over 7424.00 frames. ], tot_loss[loss=0.317, simple_loss=0.3701, pruned_loss=0.132, over 1618498.10 frames. ], batch size: 17, lr: 1.88e-02, grad_scale: 8.0 +2023-02-05 23:03:14,379 INFO [optim.py:369] (0/4) 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 23:03:24,275 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26335.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 23:03:34,742 INFO [train.py:901] (0/4) Epoch 4, batch 2100, loss[loss=0.2446, simple_loss=0.3033, pruned_loss=0.0929, over 7234.00 frames. ], tot_loss[loss=0.3143, simple_loss=0.368, pruned_loss=0.1303, over 1615312.77 frames. ], batch size: 16, lr: 1.88e-02, grad_scale: 8.0 +2023-02-05 23:04:08,222 INFO [train.py:901] (0/4) Epoch 4, batch 2150, loss[loss=0.3357, simple_loss=0.3989, pruned_loss=0.1362, over 8595.00 frames. ], tot_loss[loss=0.3141, simple_loss=0.368, pruned_loss=0.1301, over 1616672.10 frames. ], batch size: 31, lr: 1.88e-02, grad_scale: 8.0 +2023-02-05 23:04:24,393 INFO [optim.py:369] (0/4) 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 23:04:31,149 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 4, batch 2200, loss[loss=0.3557, simple_loss=0.3941, pruned_loss=0.1587, over 8580.00 frames. ], tot_loss[loss=0.3127, simple_loss=0.3668, pruned_loss=0.1293, over 1617289.32 frames. ], batch size: 39, lr: 1.88e-02, grad_scale: 8.0 +2023-02-05 23:04:50,514 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6304, 1.9575, 3.1544, 1.1533, 2.3820, 1.8697, 1.5873, 1.9369], + device='cuda:0'), covar=tensor([0.1136, 0.1265, 0.0429, 0.2578, 0.0964, 0.1774, 0.1116, 0.1564], + device='cuda:0'), in_proj_covar=tensor([0.0452, 0.0414, 0.0506, 0.0508, 0.0550, 0.0488, 0.0437, 0.0559], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:0') +2023-02-05 23:04:53,172 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=26464.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:05:18,116 INFO [train.py:901] (0/4) Epoch 4, batch 2250, loss[loss=0.3253, simple_loss=0.3827, pruned_loss=0.134, over 8360.00 frames. ], tot_loss[loss=0.3094, simple_loss=0.3643, pruned_loss=0.1273, over 1614718.16 frames. ], batch size: 24, lr: 1.88e-02, grad_scale: 8.0 +2023-02-05 23:05:33,091 INFO [optim.py:369] (0/4) 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 23:05:35,873 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.4297, 1.2661, 4.6050, 1.8053, 3.8995, 3.8235, 4.1846, 3.9713], + device='cuda:0'), covar=tensor([0.0339, 0.3407, 0.0243, 0.1962, 0.0931, 0.0480, 0.0344, 0.0474], + device='cuda:0'), in_proj_covar=tensor([0.0277, 0.0446, 0.0345, 0.0361, 0.0432, 0.0360, 0.0347, 0.0385], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-05 23:05:38,937 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 4, batch 2300, loss[loss=0.2722, simple_loss=0.3418, pruned_loss=0.1013, over 7813.00 frames. ], tot_loss[loss=0.3104, simple_loss=0.365, pruned_loss=0.1279, over 1615504.07 frames. ], batch size: 20, lr: 1.87e-02, grad_scale: 8.0 +2023-02-05 23:06:12,944 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26579.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:06:21,879 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26591.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 23:06:27,671 INFO [train.py:901] (0/4) Epoch 4, batch 2350, loss[loss=0.3024, simple_loss=0.3515, pruned_loss=0.1267, over 7285.00 frames. ], tot_loss[loss=0.3129, simple_loss=0.3673, pruned_loss=0.1292, over 1620412.48 frames. ], batch size: 16, lr: 1.87e-02, grad_scale: 8.0 +2023-02-05 23:06:35,933 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=26612.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:06:38,730 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26616.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 23:06:42,411 INFO [optim.py:369] (0/4) 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 23:06:58,649 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26645.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:06:59,875 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5035, 2.8720, 1.6894, 2.0651, 2.3428, 1.5219, 2.1080, 2.1419], + device='cuda:0'), covar=tensor([0.1260, 0.0356, 0.0921, 0.0683, 0.0555, 0.1126, 0.0813, 0.0744], + device='cuda:0'), in_proj_covar=tensor([0.0355, 0.0241, 0.0316, 0.0311, 0.0322, 0.0311, 0.0336, 0.0310], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-05 23:07:01,511 INFO [train.py:901] (0/4) Epoch 4, batch 2400, loss[loss=0.3966, simple_loss=0.4268, pruned_loss=0.1832, over 8241.00 frames. ], tot_loss[loss=0.3146, simple_loss=0.3686, pruned_loss=0.1303, over 1618613.13 frames. ], batch size: 24, lr: 1.87e-02, grad_scale: 8.0 +2023-02-05 23:07:14,820 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.52 vs. limit=5.0 +2023-02-05 23:07:37,137 INFO [train.py:901] (0/4) Epoch 4, batch 2450, loss[loss=0.3619, simple_loss=0.4017, pruned_loss=0.161, over 8499.00 frames. ], tot_loss[loss=0.3152, simple_loss=0.3693, pruned_loss=0.1305, over 1618243.62 frames. ], batch size: 39, lr: 1.87e-02, grad_scale: 8.0 +2023-02-05 23:07:42,981 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 +2023-02-05 23:07:51,855 INFO [optim.py:369] (0/4) 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 23:07:55,328 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26727.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:08:10,589 INFO [train.py:901] (0/4) Epoch 4, batch 2500, loss[loss=0.2539, simple_loss=0.3197, pruned_loss=0.09401, over 7970.00 frames. ], tot_loss[loss=0.313, simple_loss=0.3676, pruned_loss=0.1291, over 1619258.18 frames. ], batch size: 21, lr: 1.87e-02, grad_scale: 8.0 +2023-02-05 23:08:29,254 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=26776.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:08:45,264 INFO [train.py:901] (0/4) Epoch 4, batch 2550, loss[loss=0.2797, simple_loss=0.3386, pruned_loss=0.1103, over 7540.00 frames. ], tot_loss[loss=0.312, simple_loss=0.3666, pruned_loss=0.1287, over 1617811.07 frames. ], batch size: 18, lr: 1.87e-02, grad_scale: 8.0 +2023-02-05 23:09:01,302 INFO [optim.py:369] (0/4) 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 23:09:06,365 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4898, 1.1496, 1.4224, 1.0948, 1.0749, 1.2668, 1.1689, 1.3382], + device='cuda:0'), covar=tensor([0.0728, 0.1606, 0.2429, 0.1728, 0.0719, 0.2049, 0.0906, 0.0665], + device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0199, 0.0240, 0.0201, 0.0160, 0.0203, 0.0166, 0.0166], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0006, 0.0007, 0.0007, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-05 23:09:10,581 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 4, batch 2600, loss[loss=0.3068, simple_loss=0.36, pruned_loss=0.1268, over 8285.00 frames. ], tot_loss[loss=0.3126, simple_loss=0.3667, pruned_loss=0.1292, over 1616102.24 frames. ], batch size: 23, lr: 1.86e-02, grad_scale: 8.0 +2023-02-05 23:09:27,742 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26860.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:09:41,431 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5057, 2.0096, 2.2150, 0.7425, 2.3092, 1.5434, 0.6662, 1.8290], + device='cuda:0'), covar=tensor([0.0191, 0.0089, 0.0080, 0.0172, 0.0112, 0.0285, 0.0250, 0.0078], + device='cuda:0'), in_proj_covar=tensor([0.0261, 0.0187, 0.0154, 0.0227, 0.0175, 0.0309, 0.0251, 0.0210], + device='cuda:0'), out_proj_covar=tensor([1.0952e-04, 7.7415e-05, 6.2289e-05, 9.2108e-05, 7.4025e-05, 1.3876e-04, + 1.0655e-04, 8.5655e-05], device='cuda:0') +2023-02-05 23:09:50,377 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26891.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:09:56,440 INFO [train.py:901] (0/4) Epoch 4, batch 2650, loss[loss=0.3063, simple_loss=0.3759, pruned_loss=0.1184, over 8731.00 frames. ], tot_loss[loss=0.3129, simple_loss=0.3678, pruned_loss=0.129, over 1621998.69 frames. ], batch size: 34, lr: 1.86e-02, grad_scale: 8.0 +2023-02-05 23:09:57,292 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26901.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:10:12,343 INFO [optim.py:369] (0/4) 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 23:10:12,526 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26922.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 23:10:15,347 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26926.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:10:32,094 INFO [train.py:901] (0/4) Epoch 4, batch 2700, loss[loss=0.2947, simple_loss=0.3628, pruned_loss=0.1134, over 8446.00 frames. ], tot_loss[loss=0.3139, simple_loss=0.3681, pruned_loss=0.1298, over 1619090.98 frames. ], batch size: 27, lr: 1.86e-02, grad_scale: 8.0 +2023-02-05 23:10:44,338 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26968.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:10:54,311 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26983.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:11:06,043 INFO [train.py:901] (0/4) Epoch 4, batch 2750, loss[loss=0.303, simple_loss=0.3504, pruned_loss=0.1278, over 7977.00 frames. ], tot_loss[loss=0.3145, simple_loss=0.3683, pruned_loss=0.1304, over 1618508.38 frames. ], batch size: 21, lr: 1.86e-02, grad_scale: 8.0 +2023-02-05 23:11:07,092 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 +2023-02-05 23:11:12,373 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=27008.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:11:21,393 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 4, batch 2800, loss[loss=0.3826, simple_loss=0.4091, pruned_loss=0.178, over 7107.00 frames. ], tot_loss[loss=0.3147, simple_loss=0.3684, pruned_loss=0.1305, over 1616026.55 frames. ], batch size: 71, lr: 1.86e-02, grad_scale: 8.0 +2023-02-05 23:11:53,133 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.5818, 1.9279, 3.2338, 2.7568, 2.6156, 2.0218, 1.3821, 1.2701], + device='cuda:0'), covar=tensor([0.1136, 0.1528, 0.0310, 0.0596, 0.0660, 0.0629, 0.0833, 0.1459], + device='cuda:0'), in_proj_covar=tensor([0.0667, 0.0596, 0.0498, 0.0559, 0.0678, 0.0543, 0.0551, 0.0561], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 23:12:14,849 INFO [train.py:901] (0/4) Epoch 4, batch 2850, loss[loss=0.3709, simple_loss=0.4142, pruned_loss=0.1638, over 8196.00 frames. ], tot_loss[loss=0.3161, simple_loss=0.3693, pruned_loss=0.1314, over 1614834.72 frames. ], batch size: 23, lr: 1.86e-02, grad_scale: 8.0 +2023-02-05 23:12:30,251 INFO [optim.py:369] (0/4) 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 23:12:47,561 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 4, batch 2900, loss[loss=0.2792, simple_loss=0.3381, pruned_loss=0.1102, over 7976.00 frames. ], tot_loss[loss=0.3173, simple_loss=0.37, pruned_loss=0.1323, over 1614193.18 frames. ], batch size: 21, lr: 1.85e-02, grad_scale: 8.0 +2023-02-05 23:13:00,613 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=27166.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:13:05,273 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=27172.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:13:11,802 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-05 23:13:24,193 INFO [train.py:901] (0/4) Epoch 4, batch 2950, loss[loss=0.2551, simple_loss=0.3314, pruned_loss=0.08938, over 8112.00 frames. ], tot_loss[loss=0.3177, simple_loss=0.3706, pruned_loss=0.1324, over 1615551.55 frames. ], batch size: 23, lr: 1.85e-02, grad_scale: 8.0 +2023-02-05 23:13:38,737 INFO [optim.py:369] (0/4) 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 23:13:58,840 INFO [train.py:901] (0/4) Epoch 4, batch 3000, loss[loss=0.4059, simple_loss=0.4328, pruned_loss=0.1894, over 8529.00 frames. ], tot_loss[loss=0.3136, simple_loss=0.3678, pruned_loss=0.1297, over 1616361.02 frames. ], batch size: 49, lr: 1.85e-02, grad_scale: 8.0 +2023-02-05 23:13:58,840 INFO [train.py:926] (0/4) Computing validation loss +2023-02-05 23:14:05,912 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6665, 1.8439, 1.4963, 2.1924, 1.3166, 1.2971, 1.5652, 1.8906], + device='cuda:0'), covar=tensor([0.1049, 0.1164, 0.1676, 0.0681, 0.1626, 0.2223, 0.1424, 0.1068], + device='cuda:0'), in_proj_covar=tensor([0.0283, 0.0288, 0.0306, 0.0225, 0.0268, 0.0295, 0.0307, 0.0276], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0004, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-05 23:14:11,268 INFO [train.py:935] (0/4) Epoch 4, validation: loss=0.2374, simple_loss=0.3304, pruned_loss=0.07225, over 944034.00 frames. +2023-02-05 23:14:11,268 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6689MB +2023-02-05 23:14:23,004 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=27266.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 23:14:31,389 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 +2023-02-05 23:14:45,719 INFO [train.py:901] (0/4) Epoch 4, batch 3050, loss[loss=0.2767, simple_loss=0.3388, pruned_loss=0.1074, over 8247.00 frames. ], tot_loss[loss=0.3135, simple_loss=0.3679, pruned_loss=0.1295, over 1621353.84 frames. ], batch size: 22, lr: 1.85e-02, grad_scale: 8.0 +2023-02-05 23:14:54,658 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=27312.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:15:01,945 INFO [optim.py:369] (0/4) 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 23:15:06,198 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4678, 1.8171, 3.1506, 1.0839, 2.2279, 1.6757, 1.5508, 1.7476], + device='cuda:0'), covar=tensor([0.1442, 0.1728, 0.0566, 0.2851, 0.1263, 0.2268, 0.1239, 0.2044], + device='cuda:0'), in_proj_covar=tensor([0.0452, 0.0422, 0.0503, 0.0510, 0.0561, 0.0492, 0.0432, 0.0566], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:0') +2023-02-05 23:15:20,639 INFO [train.py:901] (0/4) Epoch 4, batch 3100, loss[loss=0.3292, simple_loss=0.3819, pruned_loss=0.1383, over 8606.00 frames. ], tot_loss[loss=0.3145, simple_loss=0.3684, pruned_loss=0.1303, over 1616901.27 frames. ], batch size: 31, lr: 1.85e-02, grad_scale: 8.0 +2023-02-05 23:15:41,818 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=27381.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 23:15:54,967 INFO [train.py:901] (0/4) Epoch 4, batch 3150, loss[loss=0.2584, simple_loss=0.317, pruned_loss=0.09993, over 7815.00 frames. ], tot_loss[loss=0.3144, simple_loss=0.3682, pruned_loss=0.1303, over 1615903.92 frames. ], batch size: 20, lr: 1.85e-02, grad_scale: 8.0 +2023-02-05 23:16:09,488 INFO [optim.py:369] (0/4) 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 23:16:13,654 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=27427.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:16:29,608 INFO [train.py:901] (0/4) Epoch 4, batch 3200, loss[loss=0.3078, simple_loss=0.3478, pruned_loss=0.1339, over 7808.00 frames. ], tot_loss[loss=0.3129, simple_loss=0.3672, pruned_loss=0.1293, over 1612742.18 frames. ], batch size: 19, lr: 1.84e-02, grad_scale: 8.0 +2023-02-05 23:17:03,119 INFO [train.py:901] (0/4) Epoch 4, batch 3250, loss[loss=0.3763, simple_loss=0.4247, pruned_loss=0.164, over 8130.00 frames. ], tot_loss[loss=0.3135, simple_loss=0.3676, pruned_loss=0.1297, over 1616194.49 frames. ], batch size: 22, lr: 1.84e-02, grad_scale: 8.0 +2023-02-05 23:17:10,576 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=27510.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:17:18,427 INFO [optim.py:369] (0/4) 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 23:17:19,284 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3957, 2.5138, 1.5702, 2.2140, 1.9043, 1.3722, 1.6648, 1.9869], + device='cuda:0'), covar=tensor([0.1005, 0.0293, 0.0809, 0.0506, 0.0588, 0.1013, 0.0922, 0.0616], + device='cuda:0'), in_proj_covar=tensor([0.0347, 0.0239, 0.0308, 0.0299, 0.0322, 0.0307, 0.0332, 0.0305], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-05 23:17:26,726 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3188, 1.6952, 1.6367, 0.6866, 1.6602, 1.2813, 0.2747, 1.6125], + device='cuda:0'), covar=tensor([0.0152, 0.0088, 0.0090, 0.0144, 0.0112, 0.0277, 0.0250, 0.0064], + device='cuda:0'), in_proj_covar=tensor([0.0266, 0.0194, 0.0159, 0.0230, 0.0183, 0.0313, 0.0253, 0.0213], + device='cuda:0'), out_proj_covar=tensor([1.1027e-04, 7.9632e-05, 6.3276e-05, 9.1527e-05, 7.6609e-05, 1.3842e-04, + 1.0639e-04, 8.5385e-05], device='cuda:0') +2023-02-05 23:17:37,483 INFO [train.py:901] (0/4) Epoch 4, batch 3300, loss[loss=0.3615, simple_loss=0.4156, pruned_loss=0.1537, over 8523.00 frames. ], tot_loss[loss=0.312, simple_loss=0.3665, pruned_loss=0.1288, over 1616911.88 frames. ], batch size: 34, lr: 1.84e-02, grad_scale: 8.0 +2023-02-05 23:18:12,408 INFO [train.py:901] (0/4) Epoch 4, batch 3350, loss[loss=0.2725, simple_loss=0.3387, pruned_loss=0.1031, over 8307.00 frames. ], tot_loss[loss=0.3119, simple_loss=0.3662, pruned_loss=0.1289, over 1611583.10 frames. ], batch size: 23, lr: 1.84e-02, grad_scale: 8.0 +2023-02-05 23:18:28,385 INFO [optim.py:369] (0/4) 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 23:18:30,472 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=27625.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:18:38,342 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=27637.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 23:18:46,882 INFO [train.py:901] (0/4) Epoch 4, batch 3400, loss[loss=0.3356, simple_loss=0.3869, pruned_loss=0.1421, over 8297.00 frames. ], tot_loss[loss=0.3136, simple_loss=0.3673, pruned_loss=0.13, over 1609260.94 frames. ], batch size: 23, lr: 1.84e-02, grad_scale: 16.0 +2023-02-05 23:18:54,513 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9143, 2.4480, 2.8274, 1.0878, 2.7001, 1.9329, 1.3935, 1.9507], + device='cuda:0'), covar=tensor([0.0254, 0.0104, 0.0120, 0.0182, 0.0109, 0.0236, 0.0258, 0.0115], + device='cuda:0'), in_proj_covar=tensor([0.0267, 0.0195, 0.0157, 0.0227, 0.0181, 0.0312, 0.0254, 0.0216], + device='cuda:0'), out_proj_covar=tensor([1.1066e-04, 8.0368e-05, 6.2371e-05, 9.0077e-05, 7.5706e-05, 1.3771e-04, + 1.0684e-04, 8.6755e-05], device='cuda:0') +2023-02-05 23:18:55,787 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=27662.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 23:18:56,358 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6851, 2.5665, 1.8230, 3.0679, 1.1870, 1.4164, 1.5608, 2.5319], + device='cuda:0'), covar=tensor([0.1477, 0.1206, 0.2052, 0.0567, 0.2045, 0.2622, 0.1765, 0.0997], + device='cuda:0'), in_proj_covar=tensor([0.0279, 0.0292, 0.0308, 0.0226, 0.0270, 0.0297, 0.0309, 0.0278], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0004, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-05 23:19:10,310 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=27683.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:19:21,465 INFO [train.py:901] (0/4) Epoch 4, batch 3450, loss[loss=0.3061, simple_loss=0.3745, pruned_loss=0.1189, over 8325.00 frames. ], tot_loss[loss=0.3124, simple_loss=0.3666, pruned_loss=0.1292, over 1612601.63 frames. ], batch size: 25, lr: 1.84e-02, grad_scale: 16.0 +2023-02-05 23:19:26,937 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=27708.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:19:36,056 INFO [optim.py:369] (0/4) 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 23:19:43,534 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=27732.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 23:19:55,901 INFO [train.py:901] (0/4) Epoch 4, batch 3500, loss[loss=0.3862, simple_loss=0.4084, pruned_loss=0.182, over 8234.00 frames. ], tot_loss[loss=0.3126, simple_loss=0.3666, pruned_loss=0.1293, over 1614711.39 frames. ], batch size: 49, lr: 1.83e-02, grad_scale: 16.0 +2023-02-05 23:20:10,685 WARNING [train.py:1067] (0/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-05 23:20:31,241 INFO [train.py:901] (0/4) Epoch 4, batch 3550, loss[loss=0.2853, simple_loss=0.3465, pruned_loss=0.112, over 8630.00 frames. ], tot_loss[loss=0.3111, simple_loss=0.3657, pruned_loss=0.1283, over 1616239.32 frames. ], batch size: 49, lr: 1.83e-02, grad_scale: 16.0 +2023-02-05 23:20:46,086 INFO [optim.py:369] (0/4) 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 23:21:05,474 INFO [train.py:901] (0/4) Epoch 4, batch 3600, loss[loss=0.2809, simple_loss=0.3463, pruned_loss=0.1078, over 8470.00 frames. ], tot_loss[loss=0.3123, simple_loss=0.3664, pruned_loss=0.1291, over 1616814.71 frames. ], batch size: 25, lr: 1.83e-02, grad_scale: 16.0 +2023-02-05 23:21:27,347 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=27881.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:21:39,931 INFO [train.py:901] (0/4) Epoch 4, batch 3650, loss[loss=0.2788, simple_loss=0.3504, pruned_loss=0.1036, over 8300.00 frames. ], tot_loss[loss=0.3112, simple_loss=0.3658, pruned_loss=0.1283, over 1614084.76 frames. ], batch size: 23, lr: 1.83e-02, grad_scale: 16.0 +2023-02-05 23:21:44,903 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=27906.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:21:56,103 INFO [optim.py:369] (0/4) 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 23:22:13,486 WARNING [train.py:1067] (0/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-05 23:22:14,780 INFO [train.py:901] (0/4) Epoch 4, batch 3700, loss[loss=0.2702, simple_loss=0.3302, pruned_loss=0.1051, over 7706.00 frames. ], tot_loss[loss=0.3118, simple_loss=0.3664, pruned_loss=0.1287, over 1617046.20 frames. ], batch size: 18, lr: 1.83e-02, grad_scale: 16.0 +2023-02-05 23:22:49,711 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-28000.pt +2023-02-05 23:22:50,647 INFO [train.py:901] (0/4) Epoch 4, batch 3750, loss[loss=0.3347, simple_loss=0.3815, pruned_loss=0.1439, over 7970.00 frames. ], tot_loss[loss=0.3115, simple_loss=0.3659, pruned_loss=0.1285, over 1612499.94 frames. ], batch size: 21, lr: 1.83e-02, grad_scale: 8.0 +2023-02-05 23:23:05,800 INFO [optim.py:369] (0/4) 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 23:23:09,663 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.66 vs. limit=2.0 +2023-02-05 23:23:23,378 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5961, 1.9060, 2.0850, 1.7413, 1.0349, 1.9822, 0.1406, 1.2312], + device='cuda:0'), covar=tensor([0.4073, 0.2906, 0.1033, 0.2284, 0.7932, 0.1123, 0.6958, 0.2606], + device='cuda:0'), in_proj_covar=tensor([0.0125, 0.0111, 0.0084, 0.0160, 0.0194, 0.0081, 0.0150, 0.0119], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 23:23:25,244 INFO [train.py:901] (0/4) Epoch 4, batch 3800, loss[loss=0.3356, simple_loss=0.3623, pruned_loss=0.1544, over 7421.00 frames. ], tot_loss[loss=0.3119, simple_loss=0.3659, pruned_loss=0.129, over 1611927.58 frames. ], batch size: 17, lr: 1.83e-02, grad_scale: 8.0 +2023-02-05 23:23:25,396 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.5007, 1.3975, 4.6866, 1.7064, 4.0227, 3.9025, 4.1040, 4.0623], + device='cuda:0'), covar=tensor([0.0415, 0.3507, 0.0304, 0.2079, 0.1030, 0.0566, 0.0471, 0.0476], + device='cuda:0'), in_proj_covar=tensor([0.0285, 0.0459, 0.0355, 0.0370, 0.0442, 0.0367, 0.0357, 0.0393], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-05 23:23:26,807 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6079, 1.9839, 2.2158, 0.7120, 2.2467, 1.4556, 0.5694, 1.7851], + device='cuda:0'), covar=tensor([0.0158, 0.0081, 0.0077, 0.0184, 0.0128, 0.0259, 0.0236, 0.0078], + device='cuda:0'), in_proj_covar=tensor([0.0264, 0.0197, 0.0156, 0.0230, 0.0185, 0.0315, 0.0255, 0.0217], + device='cuda:0'), out_proj_covar=tensor([1.0842e-04, 8.0473e-05, 6.0952e-05, 9.1517e-05, 7.6514e-05, 1.3804e-04, + 1.0640e-04, 8.6654e-05], device='cuda:0') +2023-02-05 23:23:42,794 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=28076.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 23:24:00,297 INFO [train.py:901] (0/4) Epoch 4, batch 3850, loss[loss=0.3176, simple_loss=0.3857, pruned_loss=0.1248, over 8337.00 frames. ], tot_loss[loss=0.3117, simple_loss=0.3659, pruned_loss=0.1287, over 1617595.34 frames. ], batch size: 25, lr: 1.82e-02, grad_scale: 8.0 +2023-02-05 23:24:15,217 INFO [optim.py:369] (0/4) 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 23:24:17,316 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-05 23:24:34,697 INFO [train.py:901] (0/4) Epoch 4, batch 3900, loss[loss=0.2892, simple_loss=0.3477, pruned_loss=0.1153, over 7664.00 frames. ], tot_loss[loss=0.3114, simple_loss=0.3662, pruned_loss=0.1283, over 1620926.04 frames. ], batch size: 19, lr: 1.82e-02, grad_scale: 8.0 +2023-02-05 23:24:43,612 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3529, 1.4548, 1.5653, 0.8970, 1.4600, 1.1248, 0.6096, 1.4045], + device='cuda:0'), covar=tensor([0.0115, 0.0071, 0.0045, 0.0105, 0.0074, 0.0194, 0.0190, 0.0054], + device='cuda:0'), in_proj_covar=tensor([0.0263, 0.0190, 0.0155, 0.0226, 0.0184, 0.0311, 0.0250, 0.0216], + device='cuda:0'), out_proj_covar=tensor([1.0746e-04, 7.6893e-05, 6.0238e-05, 8.9438e-05, 7.5971e-05, 1.3585e-04, + 1.0406e-04, 8.6044e-05], device='cuda:0') +2023-02-05 23:25:02,700 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28191.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 23:25:08,703 INFO [train.py:901] (0/4) Epoch 4, batch 3950, loss[loss=0.3672, simple_loss=0.4144, pruned_loss=0.16, over 8467.00 frames. ], tot_loss[loss=0.3125, simple_loss=0.3668, pruned_loss=0.1291, over 1618451.83 frames. ], batch size: 39, lr: 1.82e-02, grad_scale: 8.0 +2023-02-05 23:25:24,839 INFO [optim.py:369] (0/4) 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 23:25:41,184 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28247.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:25:43,098 INFO [train.py:901] (0/4) Epoch 4, batch 4000, loss[loss=0.2897, simple_loss=0.3771, pruned_loss=0.1011, over 8356.00 frames. ], tot_loss[loss=0.312, simple_loss=0.3668, pruned_loss=0.1286, over 1617842.15 frames. ], batch size: 24, lr: 1.82e-02, grad_scale: 8.0 +2023-02-05 23:25:59,930 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28273.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 23:26:01,392 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4423, 2.0972, 3.5039, 1.0903, 2.4148, 1.6049, 1.6335, 2.0645], + device='cuda:0'), covar=tensor([0.1296, 0.1392, 0.0589, 0.2557, 0.1176, 0.2108, 0.1146, 0.1848], + device='cuda:0'), in_proj_covar=tensor([0.0460, 0.0432, 0.0514, 0.0513, 0.0577, 0.0498, 0.0439, 0.0575], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:0') +2023-02-05 23:26:17,610 INFO [train.py:901] (0/4) Epoch 4, batch 4050, loss[loss=0.3006, simple_loss=0.3538, pruned_loss=0.1238, over 8337.00 frames. ], tot_loss[loss=0.3122, simple_loss=0.3669, pruned_loss=0.1287, over 1617992.21 frames. ], batch size: 25, lr: 1.82e-02, grad_scale: 8.0 +2023-02-05 23:26:21,154 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28305.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:26:34,450 INFO [optim.py:369] (0/4) 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 23:26:47,080 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4737, 2.0205, 2.2293, 1.0143, 2.2212, 1.3579, 0.7557, 1.7190], + device='cuda:0'), covar=tensor([0.0203, 0.0085, 0.0076, 0.0168, 0.0131, 0.0280, 0.0241, 0.0089], + device='cuda:0'), in_proj_covar=tensor([0.0266, 0.0194, 0.0155, 0.0231, 0.0188, 0.0313, 0.0253, 0.0220], + device='cuda:0'), out_proj_covar=tensor([1.0823e-04, 7.8485e-05, 5.9572e-05, 9.1054e-05, 7.7504e-05, 1.3668e-04, + 1.0471e-04, 8.6932e-05], device='cuda:0') +2023-02-05 23:26:50,717 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 +2023-02-05 23:26:52,366 INFO [train.py:901] (0/4) Epoch 4, batch 4100, loss[loss=0.3202, simple_loss=0.3731, pruned_loss=0.1336, over 8142.00 frames. ], tot_loss[loss=0.3106, simple_loss=0.3656, pruned_loss=0.1278, over 1614006.99 frames. ], batch size: 22, lr: 1.82e-02, grad_scale: 8.0 +2023-02-05 23:27:27,460 INFO [train.py:901] (0/4) Epoch 4, batch 4150, loss[loss=0.3814, simple_loss=0.4106, pruned_loss=0.1761, over 8512.00 frames. ], tot_loss[loss=0.3106, simple_loss=0.3655, pruned_loss=0.1278, over 1615798.49 frames. ], batch size: 28, lr: 1.81e-02, grad_scale: 8.0 +2023-02-05 23:27:43,618 INFO [optim.py:369] (0/4) 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 23:27:56,827 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 +2023-02-05 23:28:00,676 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28447.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 23:28:02,476 INFO [train.py:901] (0/4) Epoch 4, batch 4200, loss[loss=0.3266, simple_loss=0.3884, pruned_loss=0.1324, over 8337.00 frames. ], tot_loss[loss=0.3092, simple_loss=0.3643, pruned_loss=0.127, over 1616100.94 frames. ], batch size: 25, lr: 1.81e-02, grad_scale: 8.0 +2023-02-05 23:28:07,668 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-05 23:28:17,496 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28472.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 23:28:29,058 WARNING [train.py:1067] (0/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-05 23:28:36,444 INFO [train.py:901] (0/4) Epoch 4, batch 4250, loss[loss=0.2934, simple_loss=0.3427, pruned_loss=0.122, over 7521.00 frames. ], tot_loss[loss=0.3092, simple_loss=0.3643, pruned_loss=0.127, over 1615086.64 frames. ], batch size: 18, lr: 1.81e-02, grad_scale: 8.0 +2023-02-05 23:28:39,186 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28504.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:28:43,326 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28510.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:28:51,869 INFO [optim.py:369] (0/4) 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 23:29:10,383 INFO [train.py:901] (0/4) Epoch 4, batch 4300, loss[loss=0.2867, simple_loss=0.3537, pruned_loss=0.1098, over 8468.00 frames. ], tot_loss[loss=0.3077, simple_loss=0.3634, pruned_loss=0.1261, over 1613910.68 frames. ], batch size: 25, lr: 1.81e-02, grad_scale: 8.0 +2023-02-05 23:29:38,451 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=28591.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:29:45,350 INFO [train.py:901] (0/4) Epoch 4, batch 4350, loss[loss=0.2889, simple_loss=0.3574, pruned_loss=0.1102, over 8142.00 frames. ], tot_loss[loss=0.3081, simple_loss=0.3634, pruned_loss=0.1264, over 1612337.83 frames. ], batch size: 22, lr: 1.81e-02, grad_scale: 8.0 +2023-02-05 23:29:57,577 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=28617.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 23:29:58,774 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-05 23:30:01,437 INFO [optim.py:369] (0/4) 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 23:30:19,069 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=28649.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:30:19,683 INFO [train.py:901] (0/4) Epoch 4, batch 4400, loss[loss=0.251, simple_loss=0.3165, pruned_loss=0.09274, over 7434.00 frames. ], tot_loss[loss=0.3077, simple_loss=0.3634, pruned_loss=0.126, over 1611560.17 frames. ], batch size: 17, lr: 1.81e-02, grad_scale: 8.0 +2023-02-05 23:30:41,082 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-05 23:30:54,264 INFO [train.py:901] (0/4) Epoch 4, batch 4450, loss[loss=0.2675, simple_loss=0.325, pruned_loss=0.105, over 7432.00 frames. ], tot_loss[loss=0.3068, simple_loss=0.3627, pruned_loss=0.1255, over 1610616.17 frames. ], batch size: 17, lr: 1.80e-02, grad_scale: 8.0 +2023-02-05 23:30:58,473 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28706.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:31:10,733 INFO [optim.py:369] (0/4) 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 23:31:17,718 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28732.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 23:31:30,204 INFO [train.py:901] (0/4) Epoch 4, batch 4500, loss[loss=0.2801, simple_loss=0.3556, pruned_loss=0.1023, over 8341.00 frames. ], tot_loss[loss=0.3061, simple_loss=0.3621, pruned_loss=0.125, over 1608527.23 frames. ], batch size: 26, lr: 1.80e-02, grad_scale: 8.0 +2023-02-05 23:31:36,222 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-05 23:31:39,887 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28764.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:32:05,576 INFO [train.py:901] (0/4) Epoch 4, batch 4550, loss[loss=0.2733, simple_loss=0.3225, pruned_loss=0.112, over 7554.00 frames. ], tot_loss[loss=0.3079, simple_loss=0.3639, pruned_loss=0.126, over 1610987.38 frames. ], batch size: 18, lr: 1.80e-02, grad_scale: 8.0 +2023-02-05 23:32:21,342 INFO [optim.py:369] (0/4) 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 23:32:39,606 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=28848.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:32:40,909 INFO [train.py:901] (0/4) Epoch 4, batch 4600, loss[loss=0.2966, simple_loss=0.3686, pruned_loss=0.1122, over 8483.00 frames. ], tot_loss[loss=0.3084, simple_loss=0.3643, pruned_loss=0.1263, over 1612354.50 frames. ], batch size: 49, lr: 1.80e-02, grad_scale: 8.0 +2023-02-05 23:32:43,584 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=28854.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:33:00,429 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28879.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:33:14,780 INFO [train.py:901] (0/4) Epoch 4, batch 4650, loss[loss=0.3836, simple_loss=0.4145, pruned_loss=0.1764, over 8504.00 frames. ], tot_loss[loss=0.31, simple_loss=0.365, pruned_loss=0.1275, over 1613898.47 frames. ], batch size: 29, lr: 1.80e-02, grad_scale: 8.0 +2023-02-05 23:33:19,736 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 +2023-02-05 23:33:30,678 INFO [optim.py:369] (0/4) 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 23:33:49,349 INFO [train.py:901] (0/4) Epoch 4, batch 4700, loss[loss=0.2565, simple_loss=0.3183, pruned_loss=0.09733, over 7722.00 frames. ], tot_loss[loss=0.3081, simple_loss=0.3635, pruned_loss=0.1264, over 1612728.98 frames. ], batch size: 18, lr: 1.80e-02, grad_scale: 8.0 +2023-02-05 23:33:58,481 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28962.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:33:59,185 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28963.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:34:03,890 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28969.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:34:15,883 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28987.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:34:16,546 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28988.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 23:34:24,504 INFO [train.py:901] (0/4) Epoch 4, batch 4750, loss[loss=0.3347, simple_loss=0.3946, pruned_loss=0.1374, over 8490.00 frames. ], tot_loss[loss=0.3069, simple_loss=0.3627, pruned_loss=0.1256, over 1613512.34 frames. ], batch size: 29, lr: 1.80e-02, grad_scale: 8.0 +2023-02-05 23:34:33,275 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29013.0, num_to_drop=1, layers_to_drop={1} +2023-02-05 23:34:37,313 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.5661, 1.6819, 1.5341, 1.3523, 1.3608, 1.4241, 1.7800, 1.7875], + device='cuda:0'), covar=tensor([0.0593, 0.1308, 0.1908, 0.1458, 0.0739, 0.1678, 0.0850, 0.0585], + device='cuda:0'), in_proj_covar=tensor([0.0153, 0.0197, 0.0233, 0.0197, 0.0153, 0.0202, 0.0161, 0.0165], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0007, 0.0007, 0.0007, 0.0005, 0.0007, 0.0006, 0.0005], + device='cuda:0') +2023-02-05 23:34:38,681 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29020.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:34:38,759 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-05 23:34:42,471 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-05 23:34:50,947 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 +2023-02-05 23:34:56,217 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 4, batch 4800, loss[loss=0.2793, simple_loss=0.351, pruned_loss=0.1038, over 8576.00 frames. ], tot_loss[loss=0.3075, simple_loss=0.3631, pruned_loss=0.1259, over 1613814.70 frames. ], batch size: 31, lr: 1.79e-02, grad_scale: 8.0 +2023-02-05 23:35:03,001 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 +2023-02-05 23:35:34,006 INFO [train.py:901] (0/4) Epoch 4, batch 4850, loss[loss=0.3369, simple_loss=0.3653, pruned_loss=0.1542, over 7944.00 frames. ], tot_loss[loss=0.3072, simple_loss=0.3623, pruned_loss=0.126, over 1611373.96 frames. ], batch size: 20, lr: 1.79e-02, grad_scale: 8.0 +2023-02-05 23:35:34,020 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-05 23:35:49,596 INFO [optim.py:369] (0/4) 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 23:35:58,046 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.66 vs. limit=5.0 +2023-02-05 23:36:08,349 INFO [train.py:901] (0/4) Epoch 4, batch 4900, loss[loss=0.2455, simple_loss=0.304, pruned_loss=0.09349, over 7772.00 frames. ], tot_loss[loss=0.3062, simple_loss=0.3613, pruned_loss=0.1255, over 1609289.51 frames. ], batch size: 19, lr: 1.79e-02, grad_scale: 8.0 +2023-02-05 23:36:42,074 INFO [train.py:901] (0/4) Epoch 4, batch 4950, loss[loss=0.3406, simple_loss=0.3946, pruned_loss=0.1433, over 8462.00 frames. ], tot_loss[loss=0.306, simple_loss=0.361, pruned_loss=0.1255, over 1610026.94 frames. ], batch size: 29, lr: 1.79e-02, grad_scale: 8.0 +2023-02-05 23:36:47,410 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.4409, 5.3391, 4.6955, 1.9240, 4.8184, 4.9138, 5.0505, 4.5008], + device='cuda:0'), covar=tensor([0.0700, 0.0447, 0.0839, 0.4685, 0.0719, 0.0478, 0.1045, 0.0618], + device='cuda:0'), in_proj_covar=tensor([0.0375, 0.0277, 0.0299, 0.0387, 0.0300, 0.0245, 0.0292, 0.0226], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:0') +2023-02-05 23:36:56,263 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29219.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:36:58,771 INFO [optim.py:369] (0/4) 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 23:36:58,863 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=29223.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:37:00,350 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29225.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:37:04,475 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6631, 1.9886, 2.1444, 1.7021, 1.0422, 2.2600, 0.2750, 1.0849], + device='cuda:0'), covar=tensor([0.5545, 0.1711, 0.0924, 0.2730, 0.5932, 0.0861, 0.5641, 0.2687], + device='cuda:0'), in_proj_covar=tensor([0.0124, 0.0107, 0.0080, 0.0151, 0.0191, 0.0082, 0.0137, 0.0117], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 23:37:10,461 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 +2023-02-05 23:37:12,973 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29244.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:37:16,649 INFO [train.py:901] (0/4) Epoch 4, batch 5000, loss[loss=0.3165, simple_loss=0.356, pruned_loss=0.1385, over 7910.00 frames. ], tot_loss[loss=0.3073, simple_loss=0.3618, pruned_loss=0.1264, over 1609269.24 frames. ], batch size: 20, lr: 1.79e-02, grad_scale: 8.0 +2023-02-05 23:37:16,868 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29250.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:37:19,513 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29254.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:37:51,646 INFO [train.py:901] (0/4) Epoch 4, batch 5050, loss[loss=0.2465, simple_loss=0.3197, pruned_loss=0.08669, over 7922.00 frames. ], tot_loss[loss=0.3054, simple_loss=0.361, pruned_loss=0.1249, over 1609537.78 frames. ], batch size: 20, lr: 1.79e-02, grad_scale: 8.0 +2023-02-05 23:37:54,558 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2179, 2.4015, 1.4451, 1.9303, 1.8350, 1.2412, 1.6479, 1.9481], + device='cuda:0'), covar=tensor([0.1225, 0.0259, 0.0967, 0.0503, 0.0629, 0.1180, 0.0944, 0.0708], + device='cuda:0'), in_proj_covar=tensor([0.0348, 0.0237, 0.0302, 0.0301, 0.0327, 0.0309, 0.0333, 0.0306], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-05 23:38:07,699 INFO [optim.py:369] (0/4) 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 23:38:14,940 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-05 23:38:18,448 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29338.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:38:19,212 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0894, 2.5158, 2.9463, 1.1252, 3.0883, 2.0042, 1.4286, 1.7091], + device='cuda:0'), covar=tensor([0.0222, 0.0127, 0.0112, 0.0200, 0.0102, 0.0253, 0.0273, 0.0129], + device='cuda:0'), in_proj_covar=tensor([0.0280, 0.0202, 0.0160, 0.0236, 0.0190, 0.0323, 0.0262, 0.0219], + device='cuda:0'), out_proj_covar=tensor([1.1226e-04, 8.0749e-05, 6.1088e-05, 9.1787e-05, 7.7058e-05, 1.3893e-04, + 1.0667e-04, 8.5441e-05], device='cuda:0') +2023-02-05 23:38:19,602 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 +2023-02-05 23:38:26,651 INFO [train.py:901] (0/4) Epoch 4, batch 5100, loss[loss=0.3484, simple_loss=0.3936, pruned_loss=0.1516, over 8100.00 frames. ], tot_loss[loss=0.3071, simple_loss=0.3628, pruned_loss=0.1258, over 1615660.73 frames. ], batch size: 23, lr: 1.79e-02, grad_scale: 8.0 +2023-02-05 23:38:36,213 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=29364.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:38:41,414 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3289, 1.6883, 1.9078, 0.7758, 1.9234, 1.2910, 0.5604, 1.4474], + device='cuda:0'), covar=tensor([0.0141, 0.0070, 0.0049, 0.0130, 0.0082, 0.0212, 0.0204, 0.0077], + device='cuda:0'), in_proj_covar=tensor([0.0280, 0.0202, 0.0160, 0.0239, 0.0190, 0.0326, 0.0262, 0.0221], + device='cuda:0'), out_proj_covar=tensor([1.1203e-04, 8.0644e-05, 6.1241e-05, 9.2880e-05, 7.6989e-05, 1.4033e-04, + 1.0639e-04, 8.6588e-05], device='cuda:0') +2023-02-05 23:39:00,598 INFO [train.py:901] (0/4) Epoch 4, batch 5150, loss[loss=0.3314, simple_loss=0.3725, pruned_loss=0.1452, over 6828.00 frames. ], tot_loss[loss=0.308, simple_loss=0.3634, pruned_loss=0.1263, over 1614853.61 frames. ], batch size: 71, lr: 1.78e-02, grad_scale: 8.0 +2023-02-05 23:39:16,232 INFO [optim.py:369] (0/4) 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 23:39:35,362 INFO [train.py:901] (0/4) Epoch 4, batch 5200, loss[loss=0.3225, simple_loss=0.3745, pruned_loss=0.1353, over 8502.00 frames. ], tot_loss[loss=0.3094, simple_loss=0.3643, pruned_loss=0.1272, over 1613141.07 frames. ], batch size: 28, lr: 1.78e-02, grad_scale: 8.0 +2023-02-05 23:39:54,966 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29479.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:39:54,998 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6970, 1.9172, 2.0809, 1.5793, 1.0548, 2.1937, 0.4423, 1.0965], + device='cuda:0'), covar=tensor([0.3128, 0.2053, 0.0976, 0.3089, 0.7750, 0.1021, 0.5831, 0.2812], + device='cuda:0'), in_proj_covar=tensor([0.0118, 0.0109, 0.0081, 0.0154, 0.0191, 0.0080, 0.0140, 0.0116], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 23:40:02,249 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2978, 1.1498, 1.1609, 1.0261, 1.1377, 1.1098, 1.1656, 1.1035], + device='cuda:0'), covar=tensor([0.0763, 0.1488, 0.1956, 0.1533, 0.0679, 0.1794, 0.0892, 0.0665], + device='cuda:0'), in_proj_covar=tensor([0.0155, 0.0197, 0.0234, 0.0199, 0.0155, 0.0202, 0.0164, 0.0166], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0007, 0.0007, 0.0007, 0.0005, 0.0007, 0.0006, 0.0005], + device='cuda:0') +2023-02-05 23:40:09,486 INFO [train.py:901] (0/4) Epoch 4, batch 5250, loss[loss=0.2901, simple_loss=0.3374, pruned_loss=0.1214, over 7245.00 frames. ], tot_loss[loss=0.309, simple_loss=0.3643, pruned_loss=0.1269, over 1613665.35 frames. ], batch size: 16, lr: 1.78e-02, grad_scale: 4.0 +2023-02-05 23:40:12,206 WARNING [train.py:1067] (0/4) 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] (0/4) 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 23:40:26,903 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4993, 1.8341, 3.5017, 1.0654, 2.5472, 1.6179, 1.5247, 2.0476], + device='cuda:0'), covar=tensor([0.1461, 0.1804, 0.0661, 0.2928, 0.1298, 0.2402, 0.1361, 0.2290], + device='cuda:0'), in_proj_covar=tensor([0.0455, 0.0429, 0.0505, 0.0515, 0.0559, 0.0498, 0.0435, 0.0575], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:0') +2023-02-05 23:40:43,421 INFO [train.py:901] (0/4) Epoch 4, batch 5300, loss[loss=0.4092, simple_loss=0.4268, pruned_loss=0.1958, over 6552.00 frames. ], tot_loss[loss=0.3089, simple_loss=0.3643, pruned_loss=0.1268, over 1616223.94 frames. ], batch size: 71, lr: 1.78e-02, grad_scale: 4.0 +2023-02-05 23:40:57,669 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29569.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:41:14,739 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29594.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:41:17,287 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=29598.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:41:18,729 INFO [train.py:901] (0/4) Epoch 4, batch 5350, loss[loss=0.2818, simple_loss=0.3392, pruned_loss=0.1122, over 7982.00 frames. ], tot_loss[loss=0.3081, simple_loss=0.3636, pruned_loss=0.1263, over 1613358.45 frames. ], batch size: 21, lr: 1.78e-02, grad_scale: 4.0 +2023-02-05 23:41:32,360 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 23:41:53,620 INFO [train.py:901] (0/4) Epoch 4, batch 5400, loss[loss=0.2987, simple_loss=0.3743, pruned_loss=0.1115, over 8491.00 frames. ], tot_loss[loss=0.3056, simple_loss=0.3617, pruned_loss=0.1247, over 1617169.82 frames. ], batch size: 28, lr: 1.78e-02, grad_scale: 4.0 +2023-02-05 23:42:14,106 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 +2023-02-05 23:42:14,533 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29680.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:42:28,607 INFO [train.py:901] (0/4) Epoch 4, batch 5450, loss[loss=0.333, simple_loss=0.3902, pruned_loss=0.1379, over 8584.00 frames. ], tot_loss[loss=0.3052, simple_loss=0.3614, pruned_loss=0.1245, over 1614789.75 frames. ], batch size: 31, lr: 1.78e-02, grad_scale: 4.0 +2023-02-05 23:42:36,132 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9868, 1.5088, 1.4506, 1.2642, 1.3427, 1.3965, 1.5717, 1.4903], + device='cuda:0'), covar=tensor([0.0598, 0.1130, 0.1643, 0.1365, 0.0653, 0.1505, 0.0755, 0.0588], + device='cuda:0'), in_proj_covar=tensor([0.0153, 0.0196, 0.0232, 0.0197, 0.0153, 0.0200, 0.0162, 0.0165], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0007, 0.0007, 0.0007, 0.0005, 0.0007, 0.0006, 0.0005], + device='cuda:0') +2023-02-05 23:42:37,304 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29713.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:42:44,142 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 +2023-02-05 23:42:44,948 INFO [optim.py:369] (0/4) 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 23:42:52,681 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29735.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:42:58,016 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-05 23:43:02,823 INFO [train.py:901] (0/4) Epoch 4, batch 5500, loss[loss=0.3213, simple_loss=0.3892, pruned_loss=0.1267, over 8596.00 frames. ], tot_loss[loss=0.3051, simple_loss=0.3614, pruned_loss=0.1244, over 1615553.36 frames. ], batch size: 34, lr: 1.77e-02, grad_scale: 4.0 +2023-02-05 23:43:10,325 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29760.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:43:32,518 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3936, 1.2990, 2.7359, 1.1693, 2.0711, 3.0441, 2.7930, 2.5387], + device='cuda:0'), covar=tensor([0.1241, 0.1658, 0.0446, 0.2169, 0.0707, 0.0320, 0.0493, 0.0796], + device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0266, 0.0216, 0.0264, 0.0217, 0.0190, 0.0199, 0.0264], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:0') +2023-02-05 23:43:33,851 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.6632, 3.5911, 3.3121, 1.5812, 3.2448, 3.1120, 3.3757, 2.7308], + device='cuda:0'), covar=tensor([0.1090, 0.0747, 0.0959, 0.4717, 0.0852, 0.1049, 0.1411, 0.1151], + device='cuda:0'), in_proj_covar=tensor([0.0374, 0.0269, 0.0295, 0.0390, 0.0296, 0.0246, 0.0289, 0.0221], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:0') +2023-02-05 23:43:37,890 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1577, 1.7733, 4.3511, 1.7603, 2.4059, 5.0174, 4.5254, 4.4064], + device='cuda:0'), covar=tensor([0.1216, 0.1489, 0.0273, 0.1986, 0.0760, 0.0174, 0.0337, 0.0572], + device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0266, 0.0215, 0.0264, 0.0216, 0.0189, 0.0198, 0.0263], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:0') +2023-02-05 23:43:38,486 INFO [train.py:901] (0/4) Epoch 4, batch 5550, loss[loss=0.3341, simple_loss=0.3972, pruned_loss=0.1355, over 8193.00 frames. ], tot_loss[loss=0.3074, simple_loss=0.3634, pruned_loss=0.1257, over 1614138.36 frames. ], batch size: 23, lr: 1.77e-02, grad_scale: 4.0 +2023-02-05 23:43:48,539 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0874, 1.0774, 1.0676, 0.9129, 0.7622, 1.2364, 0.0154, 0.8423], + device='cuda:0'), covar=tensor([0.3639, 0.2512, 0.1293, 0.2383, 0.6249, 0.0960, 0.5758, 0.2888], + device='cuda:0'), in_proj_covar=tensor([0.0119, 0.0111, 0.0082, 0.0157, 0.0191, 0.0080, 0.0141, 0.0120], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 23:43:51,762 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) Epoch 4, batch 5600, loss[loss=0.3443, simple_loss=0.3978, pruned_loss=0.1454, over 8508.00 frames. ], tot_loss[loss=0.3073, simple_loss=0.3631, pruned_loss=0.1258, over 1609835.25 frames. ], batch size: 26, lr: 1.77e-02, grad_scale: 8.0 +2023-02-05 23:44:46,052 INFO [train.py:901] (0/4) Epoch 4, batch 5650, loss[loss=0.3182, simple_loss=0.3779, pruned_loss=0.1292, over 8247.00 frames. ], tot_loss[loss=0.3069, simple_loss=0.3625, pruned_loss=0.1257, over 1608019.41 frames. ], batch size: 24, lr: 1.77e-02, grad_scale: 8.0 +2023-02-05 23:44:55,342 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=29913.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:45:03,286 INFO [optim.py:369] (0/4) 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 23:45:03,319 WARNING [train.py:1067] (0/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-05 23:45:20,770 INFO [train.py:901] (0/4) Epoch 4, batch 5700, loss[loss=0.3122, simple_loss=0.371, pruned_loss=0.1267, over 8508.00 frames. ], tot_loss[loss=0.3052, simple_loss=0.3615, pruned_loss=0.1245, over 1606744.42 frames. ], batch size: 26, lr: 1.77e-02, grad_scale: 8.0 +2023-02-05 23:45:34,465 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29969.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:45:52,043 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29994.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:45:56,066 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-30000.pt +2023-02-05 23:45:56,992 INFO [train.py:901] (0/4) Epoch 4, batch 5750, loss[loss=0.3066, simple_loss=0.3613, pruned_loss=0.1259, over 8025.00 frames. ], tot_loss[loss=0.3026, simple_loss=0.3597, pruned_loss=0.1227, over 1611533.63 frames. ], batch size: 22, lr: 1.77e-02, grad_scale: 8.0 +2023-02-05 23:46:00,485 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.8412, 2.2929, 3.7306, 3.0056, 2.9227, 2.1620, 1.5890, 1.6710], + device='cuda:0'), covar=tensor([0.1301, 0.1926, 0.0377, 0.0785, 0.0849, 0.0809, 0.0972, 0.1863], + device='cuda:0'), in_proj_covar=tensor([0.0694, 0.0624, 0.0531, 0.0599, 0.0714, 0.0583, 0.0570, 0.0588], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 23:46:07,156 WARNING [train.py:1067] (0/4) 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] (0/4) 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 23:46:13,352 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=30024.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:46:16,813 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30028.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:46:32,303 INFO [train.py:901] (0/4) Epoch 4, batch 5800, loss[loss=0.355, simple_loss=0.4069, pruned_loss=0.1516, over 8346.00 frames. ], tot_loss[loss=0.303, simple_loss=0.3597, pruned_loss=0.1231, over 1614513.76 frames. ], batch size: 26, lr: 1.77e-02, grad_scale: 8.0 +2023-02-05 23:47:06,537 INFO [train.py:901] (0/4) Epoch 4, batch 5850, loss[loss=0.2573, simple_loss=0.3072, pruned_loss=0.1037, over 7286.00 frames. ], tot_loss[loss=0.3036, simple_loss=0.3602, pruned_loss=0.1235, over 1618933.58 frames. ], batch size: 16, lr: 1.76e-02, grad_scale: 8.0 +2023-02-05 23:47:23,103 INFO [optim.py:369] (0/4) 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 23:47:33,290 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30139.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:47:35,876 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 4, batch 5900, loss[loss=0.2966, simple_loss=0.3677, pruned_loss=0.1127, over 8321.00 frames. ], tot_loss[loss=0.3056, simple_loss=0.362, pruned_loss=0.1246, over 1618104.11 frames. ], batch size: 25, lr: 1.76e-02, grad_scale: 8.0 +2023-02-05 23:47:51,359 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=30164.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:48:16,092 INFO [train.py:901] (0/4) Epoch 4, batch 5950, loss[loss=0.2778, simple_loss=0.341, pruned_loss=0.1073, over 7645.00 frames. ], tot_loss[loss=0.3057, simple_loss=0.3617, pruned_loss=0.1248, over 1614044.72 frames. ], batch size: 19, lr: 1.76e-02, grad_scale: 8.0 +2023-02-05 23:48:24,327 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1138, 1.8683, 2.2458, 2.0956, 1.7212, 2.1944, 1.2727, 1.8671], + device='cuda:0'), covar=tensor([0.2693, 0.3093, 0.1126, 0.1780, 0.4172, 0.0873, 0.4422, 0.2377], + device='cuda:0'), in_proj_covar=tensor([0.0124, 0.0119, 0.0085, 0.0160, 0.0199, 0.0084, 0.0148, 0.0124], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 23:48:32,441 INFO [optim.py:369] (0/4) 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 23:48:35,337 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30227.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:48:50,184 INFO [train.py:901] (0/4) Epoch 4, batch 6000, loss[loss=0.3083, simple_loss=0.3682, pruned_loss=0.1242, over 8354.00 frames. ], tot_loss[loss=0.3067, simple_loss=0.3628, pruned_loss=0.1253, over 1612815.58 frames. ], batch size: 24, lr: 1.76e-02, grad_scale: 8.0 +2023-02-05 23:48:50,185 INFO [train.py:926] (0/4) Computing validation loss +2023-02-05 23:49:02,861 INFO [train.py:935] (0/4) Epoch 4, validation: loss=0.2338, simple_loss=0.3275, pruned_loss=0.07005, over 944034.00 frames. +2023-02-05 23:49:02,862 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6689MB +2023-02-05 23:49:08,402 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.6531, 1.4270, 3.7782, 1.5369, 3.2242, 3.1636, 3.3043, 3.1876], + device='cuda:0'), covar=tensor([0.0468, 0.3195, 0.0387, 0.2228, 0.1141, 0.0753, 0.0506, 0.0644], + device='cuda:0'), in_proj_covar=tensor([0.0299, 0.0461, 0.0365, 0.0377, 0.0450, 0.0380, 0.0365, 0.0407], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-05 23:49:09,870 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.2188, 2.3754, 1.8387, 1.7191, 1.8446, 2.0343, 2.3304, 2.0614], + device='cuda:0'), covar=tensor([0.0648, 0.1135, 0.1762, 0.1443, 0.0715, 0.1406, 0.0788, 0.0617], + device='cuda:0'), in_proj_covar=tensor([0.0153, 0.0196, 0.0233, 0.0198, 0.0154, 0.0201, 0.0162, 0.0165], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0007, 0.0008, 0.0007, 0.0006, 0.0007, 0.0006, 0.0005], + device='cuda:0') +2023-02-05 23:49:17,528 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 +2023-02-05 23:49:22,561 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30284.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:49:33,988 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.9375, 2.0262, 3.7599, 3.2443, 3.1783, 2.3103, 1.4354, 1.4645], + device='cuda:0'), covar=tensor([0.1474, 0.2364, 0.0452, 0.0767, 0.0891, 0.0851, 0.1072, 0.2186], + device='cuda:0'), in_proj_covar=tensor([0.0698, 0.0635, 0.0535, 0.0597, 0.0718, 0.0592, 0.0572, 0.0586], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 23:49:37,706 INFO [train.py:901] (0/4) Epoch 4, batch 6050, loss[loss=0.3414, simple_loss=0.3919, pruned_loss=0.1455, over 8641.00 frames. ], tot_loss[loss=0.3068, simple_loss=0.3626, pruned_loss=0.1255, over 1614866.73 frames. ], batch size: 49, lr: 1.76e-02, grad_scale: 8.0 +2023-02-05 23:49:44,064 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30309.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:49:47,537 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=2.03 vs. limit=2.0 +2023-02-05 23:49:53,937 INFO [optim.py:369] (0/4) 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 23:50:12,474 INFO [train.py:901] (0/4) Epoch 4, batch 6100, loss[loss=0.316, simple_loss=0.3737, pruned_loss=0.1292, over 8252.00 frames. ], tot_loss[loss=0.3052, simple_loss=0.3612, pruned_loss=0.1246, over 1615354.31 frames. ], batch size: 22, lr: 1.76e-02, grad_scale: 8.0 +2023-02-05 23:50:14,283 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.11 vs. limit=5.0 +2023-02-05 23:50:32,435 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30378.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:50:39,276 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-05 23:50:44,144 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30395.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:50:47,443 INFO [train.py:901] (0/4) Epoch 4, batch 6150, loss[loss=0.4465, simple_loss=0.469, pruned_loss=0.212, over 8612.00 frames. ], tot_loss[loss=0.3052, simple_loss=0.3615, pruned_loss=0.1245, over 1614864.13 frames. ], batch size: 31, lr: 1.76e-02, grad_scale: 8.0 +2023-02-05 23:51:02,482 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30420.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:51:05,079 INFO [optim.py:369] (0/4) 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 23:51:23,142 INFO [train.py:901] (0/4) Epoch 4, batch 6200, loss[loss=0.2494, simple_loss=0.3037, pruned_loss=0.09759, over 6785.00 frames. ], tot_loss[loss=0.3052, simple_loss=0.3613, pruned_loss=0.1246, over 1612385.12 frames. ], batch size: 15, lr: 1.75e-02, grad_scale: 8.0 +2023-02-05 23:51:48,270 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=30487.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:51:57,466 INFO [train.py:901] (0/4) Epoch 4, batch 6250, loss[loss=0.3117, simple_loss=0.3726, pruned_loss=0.1254, over 8463.00 frames. ], tot_loss[loss=0.3061, simple_loss=0.362, pruned_loss=0.1251, over 1613262.07 frames. ], batch size: 27, lr: 1.75e-02, grad_scale: 8.0 +2023-02-05 23:52:14,465 INFO [optim.py:369] (0/4) 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 23:52:22,898 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30535.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:52:32,695 INFO [train.py:901] (0/4) Epoch 4, batch 6300, loss[loss=0.263, simple_loss=0.3372, pruned_loss=0.09444, over 8256.00 frames. ], tot_loss[loss=0.3031, simple_loss=0.36, pruned_loss=0.1231, over 1614806.39 frames. ], batch size: 24, lr: 1.75e-02, grad_scale: 8.0 +2023-02-05 23:52:39,494 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=30571.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:53:06,881 INFO [train.py:901] (0/4) Epoch 4, batch 6350, loss[loss=0.2902, simple_loss=0.361, pruned_loss=0.1097, over 8287.00 frames. ], tot_loss[loss=0.3043, simple_loss=0.3611, pruned_loss=0.1237, over 1612068.30 frames. ], batch size: 23, lr: 1.75e-02, grad_scale: 8.0 +2023-02-05 23:53:08,352 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30602.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:53:23,785 INFO [optim.py:369] (0/4) 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 23:53:37,224 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.0631, 1.4379, 4.2225, 1.5519, 3.7072, 3.5535, 3.8524, 3.7271], + device='cuda:0'), covar=tensor([0.0410, 0.3214, 0.0451, 0.2344, 0.1067, 0.0687, 0.0402, 0.0531], + device='cuda:0'), in_proj_covar=tensor([0.0303, 0.0469, 0.0380, 0.0386, 0.0458, 0.0377, 0.0369, 0.0414], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-05 23:53:42,473 INFO [train.py:901] (0/4) Epoch 4, batch 6400, loss[loss=0.2577, simple_loss=0.3255, pruned_loss=0.09494, over 8280.00 frames. ], tot_loss[loss=0.3048, simple_loss=0.3615, pruned_loss=0.124, over 1611170.28 frames. ], batch size: 23, lr: 1.75e-02, grad_scale: 8.0 +2023-02-05 23:53:58,068 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30673.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:54:03,301 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 +2023-02-05 23:54:07,687 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30686.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:54:16,882 INFO [train.py:901] (0/4) Epoch 4, batch 6450, loss[loss=0.315, simple_loss=0.3587, pruned_loss=0.1356, over 8129.00 frames. ], tot_loss[loss=0.3052, simple_loss=0.3616, pruned_loss=0.1244, over 1611140.23 frames. ], batch size: 22, lr: 1.75e-02, grad_scale: 8.0 +2023-02-05 23:54:31,537 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=30722.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:54:31,608 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.5897, 1.5107, 4.6427, 1.9021, 4.1016, 3.9122, 4.3395, 4.1329], + device='cuda:0'), covar=tensor([0.0282, 0.3219, 0.0271, 0.2054, 0.0772, 0.0493, 0.0320, 0.0405], + device='cuda:0'), in_proj_covar=tensor([0.0303, 0.0471, 0.0379, 0.0391, 0.0459, 0.0380, 0.0372, 0.0417], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-05 23:54:32,829 INFO [optim.py:369] (0/4) 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 23:54:50,957 INFO [train.py:901] (0/4) Epoch 4, batch 6500, loss[loss=0.3154, simple_loss=0.3839, pruned_loss=0.1234, over 8614.00 frames. ], tot_loss[loss=0.3035, simple_loss=0.3604, pruned_loss=0.1233, over 1612054.22 frames. ], batch size: 34, lr: 1.75e-02, grad_scale: 8.0 +2023-02-05 23:55:07,036 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 +2023-02-05 23:55:18,158 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5469, 1.8089, 1.9685, 1.5834, 1.0496, 2.0109, 0.3069, 1.1835], + device='cuda:0'), covar=tensor([0.2794, 0.2103, 0.0849, 0.2077, 0.6414, 0.1009, 0.5643, 0.2478], + device='cuda:0'), in_proj_covar=tensor([0.0116, 0.0114, 0.0079, 0.0159, 0.0199, 0.0081, 0.0142, 0.0119], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-05 23:55:25,836 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 +2023-02-05 23:55:26,309 INFO [train.py:901] (0/4) Epoch 4, batch 6550, loss[loss=0.2461, simple_loss=0.3135, pruned_loss=0.08931, over 7696.00 frames. ], tot_loss[loss=0.3042, simple_loss=0.3608, pruned_loss=0.1237, over 1611403.59 frames. ], batch size: 18, lr: 1.74e-02, grad_scale: 8.0 +2023-02-05 23:55:42,637 INFO [optim.py:369] (0/4) 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 23:55:50,036 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-05 23:55:51,478 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30837.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:56:00,638 INFO [train.py:901] (0/4) Epoch 4, batch 6600, loss[loss=0.2403, simple_loss=0.299, pruned_loss=0.09078, over 7639.00 frames. ], tot_loss[loss=0.3034, simple_loss=0.3603, pruned_loss=0.1233, over 1612642.97 frames. ], batch size: 19, lr: 1.74e-02, grad_scale: 8.0 +2023-02-05 23:56:06,240 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30858.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:56:08,701 WARNING [train.py:1067] (0/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-05 23:56:24,220 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30883.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:56:35,342 INFO [train.py:901] (0/4) Epoch 4, batch 6650, loss[loss=0.3238, simple_loss=0.3893, pruned_loss=0.1291, over 8504.00 frames. ], tot_loss[loss=0.3024, simple_loss=0.3595, pruned_loss=0.1226, over 1614059.96 frames. ], batch size: 26, lr: 1.74e-02, grad_scale: 8.0 +2023-02-05 23:56:50,093 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 23:57:04,046 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30942.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:57:04,650 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 4, batch 6700, loss[loss=0.3947, simple_loss=0.4185, pruned_loss=0.1855, over 6638.00 frames. ], tot_loss[loss=0.3032, simple_loss=0.3601, pruned_loss=0.1232, over 1616616.53 frames. ], batch size: 71, lr: 1.74e-02, grad_scale: 8.0 +2023-02-05 23:57:10,674 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30952.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:57:21,614 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30967.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:57:45,172 INFO [train.py:901] (0/4) Epoch 4, batch 6750, loss[loss=0.3021, simple_loss=0.3646, pruned_loss=0.1198, over 8746.00 frames. ], tot_loss[loss=0.3024, simple_loss=0.3596, pruned_loss=0.1225, over 1617548.30 frames. ], batch size: 34, lr: 1.74e-02, grad_scale: 8.0 +2023-02-05 23:57:56,385 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=31017.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:58:00,816 INFO [optim.py:369] (0/4) 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 23:58:04,848 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31029.0, num_to_drop=1, layers_to_drop={0} +2023-02-05 23:58:18,930 INFO [train.py:901] (0/4) Epoch 4, batch 6800, loss[loss=0.3303, simple_loss=0.3919, pruned_loss=0.1344, over 8248.00 frames. ], tot_loss[loss=0.3035, simple_loss=0.3604, pruned_loss=0.1232, over 1617399.65 frames. ], batch size: 24, lr: 1.74e-02, grad_scale: 8.0 +2023-02-05 23:58:19,588 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-05 23:58:31,183 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0424, 1.7162, 1.2882, 1.7485, 1.3975, 1.0676, 1.3105, 1.5854], + device='cuda:0'), covar=tensor([0.0849, 0.0343, 0.0838, 0.0340, 0.0535, 0.1049, 0.0604, 0.0501], + device='cuda:0'), in_proj_covar=tensor([0.0347, 0.0239, 0.0303, 0.0298, 0.0321, 0.0311, 0.0330, 0.0304], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-05 23:58:48,707 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 4, batch 6850, loss[loss=0.2826, simple_loss=0.3617, pruned_loss=0.1018, over 8283.00 frames. ], tot_loss[loss=0.3018, simple_loss=0.3599, pruned_loss=0.1218, over 1620628.71 frames. ], batch size: 23, lr: 1.74e-02, grad_scale: 8.0 +2023-02-05 23:59:06,887 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31118.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:59:10,000 WARNING [train.py:1067] (0/4) 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] (0/4) 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 23:59:16,192 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31132.0, num_to_drop=0, layers_to_drop=set() +2023-02-05 23:59:28,758 INFO [train.py:901] (0/4) Epoch 4, batch 6900, loss[loss=0.2892, simple_loss=0.3467, pruned_loss=0.1158, over 8088.00 frames. ], tot_loss[loss=0.3028, simple_loss=0.3608, pruned_loss=0.1224, over 1621711.78 frames. ], batch size: 21, lr: 1.73e-02, grad_scale: 8.0 +2023-02-05 23:59:38,460 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.2979, 1.2949, 4.4097, 1.6041, 3.6919, 3.5998, 3.9352, 3.8428], + device='cuda:0'), covar=tensor([0.0409, 0.3464, 0.0297, 0.2313, 0.1107, 0.0651, 0.0416, 0.0519], + device='cuda:0'), in_proj_covar=tensor([0.0303, 0.0461, 0.0368, 0.0388, 0.0454, 0.0378, 0.0373, 0.0414], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-05 23:59:57,134 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 +2023-02-06 00:00:03,443 INFO [train.py:901] (0/4) Epoch 4, batch 6950, loss[loss=0.3355, simple_loss=0.3856, pruned_loss=0.1427, over 8368.00 frames. ], tot_loss[loss=0.3018, simple_loss=0.3597, pruned_loss=0.1219, over 1617548.64 frames. ], batch size: 24, lr: 1.73e-02, grad_scale: 8.0 +2023-02-06 00:00:18,138 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-06 00:00:20,068 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 4, batch 7000, loss[loss=0.3087, simple_loss=0.3562, pruned_loss=0.1306, over 7530.00 frames. ], tot_loss[loss=0.3033, simple_loss=0.3609, pruned_loss=0.1229, over 1620785.18 frames. ], batch size: 18, lr: 1.73e-02, grad_scale: 8.0 +2023-02-06 00:00:48,749 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=31265.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:01:03,269 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=31287.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:01:09,176 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 4, batch 7050, loss[loss=0.2749, simple_loss=0.3394, pruned_loss=0.1052, over 7928.00 frames. ], tot_loss[loss=0.3047, simple_loss=0.3615, pruned_loss=0.1239, over 1618198.47 frames. ], batch size: 20, lr: 1.73e-02, grad_scale: 8.0 +2023-02-06 00:01:28,375 INFO [optim.py:369] (0/4) 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 00:01:47,439 INFO [train.py:901] (0/4) Epoch 4, batch 7100, loss[loss=0.4096, simple_loss=0.424, pruned_loss=0.1975, over 7110.00 frames. ], tot_loss[loss=0.3054, simple_loss=0.3619, pruned_loss=0.1245, over 1615484.52 frames. ], batch size: 71, lr: 1.73e-02, grad_scale: 8.0 +2023-02-06 00:01:48,712 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-02-06 00:02:02,608 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=31373.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 00:02:08,038 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31380.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:02:13,363 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31388.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:02:21,165 INFO [train.py:901] (0/4) Epoch 4, batch 7150, loss[loss=0.2455, simple_loss=0.3062, pruned_loss=0.09238, over 7936.00 frames. ], tot_loss[loss=0.307, simple_loss=0.3626, pruned_loss=0.1256, over 1611946.96 frames. ], batch size: 20, lr: 1.73e-02, grad_scale: 8.0 +2023-02-06 00:02:22,726 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31411.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:02:29,903 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 00:02:39,209 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 4, batch 7200, loss[loss=0.2884, simple_loss=0.3326, pruned_loss=0.1221, over 7554.00 frames. ], tot_loss[loss=0.3056, simple_loss=0.3619, pruned_loss=0.1246, over 1613235.57 frames. ], batch size: 18, lr: 1.73e-02, grad_scale: 8.0 +2023-02-06 00:03:22,241 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31488.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 00:03:30,967 INFO [train.py:901] (0/4) Epoch 4, batch 7250, loss[loss=0.2765, simple_loss=0.3216, pruned_loss=0.1157, over 7533.00 frames. ], tot_loss[loss=0.3029, simple_loss=0.3592, pruned_loss=0.1233, over 1607921.39 frames. ], batch size: 18, lr: 1.73e-02, grad_scale: 16.0 +2023-02-06 00:03:37,358 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31509.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 00:03:43,637 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.63 vs. limit=5.0 +2023-02-06 00:03:47,340 INFO [optim.py:369] (0/4) 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 00:03:52,988 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4393, 1.2956, 1.4054, 1.0733, 1.1373, 1.3553, 1.2431, 1.2357], + device='cuda:0'), covar=tensor([0.0693, 0.1247, 0.1849, 0.1537, 0.0559, 0.1484, 0.0760, 0.0577], + device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0190, 0.0229, 0.0192, 0.0145, 0.0193, 0.0157, 0.0158], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0007, 0.0008, 0.0007, 0.0005, 0.0007, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 00:04:05,055 INFO [train.py:901] (0/4) Epoch 4, batch 7300, loss[loss=0.3451, simple_loss=0.4035, pruned_loss=0.1433, over 8470.00 frames. ], tot_loss[loss=0.3036, simple_loss=0.3598, pruned_loss=0.1237, over 1608998.87 frames. ], batch size: 25, lr: 1.72e-02, grad_scale: 16.0 +2023-02-06 00:04:33,536 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 +2023-02-06 00:04:40,541 INFO [train.py:901] (0/4) Epoch 4, batch 7350, loss[loss=0.2934, simple_loss=0.3599, pruned_loss=0.1135, over 8344.00 frames. ], tot_loss[loss=0.3026, simple_loss=0.3593, pruned_loss=0.123, over 1611329.52 frames. ], batch size: 26, lr: 1.72e-02, grad_scale: 16.0 +2023-02-06 00:04:57,283 INFO [optim.py:369] (0/4) 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 00:04:59,980 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-06 00:05:05,400 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31636.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:05:14,568 INFO [train.py:901] (0/4) Epoch 4, batch 7400, loss[loss=0.2602, simple_loss=0.3278, pruned_loss=0.09625, over 8245.00 frames. ], tot_loss[loss=0.3019, simple_loss=0.3589, pruned_loss=0.1224, over 1611937.39 frames. ], batch size: 24, lr: 1.72e-02, grad_scale: 16.0 +2023-02-06 00:05:19,258 WARNING [train.py:1067] (0/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-06 00:05:20,106 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31658.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:05:21,981 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31661.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:05:26,687 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31667.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:05:37,005 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31683.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:05:43,497 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 4, batch 7450, loss[loss=0.3087, simple_loss=0.3471, pruned_loss=0.1351, over 7708.00 frames. ], tot_loss[loss=0.3004, simple_loss=0.3576, pruned_loss=0.1216, over 1608188.68 frames. ], batch size: 18, lr: 1.72e-02, grad_scale: 16.0 +2023-02-06 00:05:58,000 WARNING [train.py:1067] (0/4) 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] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31722.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:06:05,434 INFO [optim.py:369] (0/4) 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 00:06:19,833 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31744.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 00:06:21,846 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4252, 1.6096, 1.6029, 1.2649, 1.4646, 1.5708, 1.8734, 1.9692], + device='cuda:0'), covar=tensor([0.0610, 0.1245, 0.1806, 0.1551, 0.0672, 0.1561, 0.0714, 0.0513], + device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0190, 0.0232, 0.0194, 0.0145, 0.0196, 0.0156, 0.0158], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0007, 0.0008, 0.0007, 0.0005, 0.0007, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 00:06:23,707 INFO [train.py:901] (0/4) Epoch 4, batch 7500, loss[loss=0.3585, simple_loss=0.4044, pruned_loss=0.1563, over 8326.00 frames. ], tot_loss[loss=0.2998, simple_loss=0.3572, pruned_loss=0.1212, over 1606300.84 frames. ], batch size: 25, lr: 1.72e-02, grad_scale: 16.0 +2023-02-06 00:06:36,772 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31769.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 00:06:37,944 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=31771.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:06:47,164 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.83 vs. limit=5.0 +2023-02-06 00:06:57,684 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0301, 1.6396, 2.3137, 2.0851, 2.1439, 1.7721, 1.3688, 0.7144], + device='cuda:0'), covar=tensor([0.1291, 0.1413, 0.0397, 0.0727, 0.0571, 0.0803, 0.0864, 0.1362], + device='cuda:0'), in_proj_covar=tensor([0.0708, 0.0632, 0.0533, 0.0609, 0.0724, 0.0592, 0.0575, 0.0588], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 00:06:58,254 INFO [train.py:901] (0/4) Epoch 4, batch 7550, loss[loss=0.2999, simple_loss=0.3377, pruned_loss=0.1311, over 7427.00 frames. ], tot_loss[loss=0.3004, simple_loss=0.3575, pruned_loss=0.1217, over 1609979.53 frames. ], batch size: 17, lr: 1.72e-02, grad_scale: 8.0 +2023-02-06 00:07:02,941 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31806.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:07:16,189 INFO [optim.py:369] (0/4) 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 00:07:27,216 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31841.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:07:27,983 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4160, 1.6810, 1.4523, 1.2420, 1.4045, 1.4044, 1.9245, 1.9223], + device='cuda:0'), covar=tensor([0.0605, 0.1210, 0.1884, 0.1598, 0.0676, 0.1599, 0.0730, 0.0540], + device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0189, 0.0233, 0.0194, 0.0144, 0.0196, 0.0156, 0.0158], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0007, 0.0008, 0.0007, 0.0005, 0.0007, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 00:07:34,061 INFO [train.py:901] (0/4) Epoch 4, batch 7600, loss[loss=0.2361, simple_loss=0.3164, pruned_loss=0.07787, over 7923.00 frames. ], tot_loss[loss=0.2986, simple_loss=0.3559, pruned_loss=0.1207, over 1605227.83 frames. ], batch size: 20, lr: 1.72e-02, grad_scale: 8.0 +2023-02-06 00:07:36,149 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=31853.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 00:07:40,185 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6059, 1.8505, 2.0597, 1.4698, 0.8905, 1.9868, 0.2979, 1.1846], + device='cuda:0'), covar=tensor([0.2884, 0.2039, 0.1591, 0.3040, 0.7070, 0.0981, 0.6173, 0.2478], + device='cuda:0'), in_proj_covar=tensor([0.0122, 0.0113, 0.0083, 0.0165, 0.0205, 0.0083, 0.0151, 0.0121], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 00:07:58,616 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31886.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:07:59,322 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3338, 1.5441, 1.5717, 1.3111, 0.8402, 1.5621, 0.1098, 1.0016], + device='cuda:0'), covar=tensor([0.4332, 0.2317, 0.1522, 0.2816, 0.7206, 0.1062, 0.6892, 0.2773], + device='cuda:0'), in_proj_covar=tensor([0.0124, 0.0113, 0.0083, 0.0165, 0.0205, 0.0083, 0.0151, 0.0122], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 00:08:05,400 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31896.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:08:07,993 INFO [train.py:901] (0/4) Epoch 4, batch 7650, loss[loss=0.3173, simple_loss=0.3755, pruned_loss=0.1296, over 8316.00 frames. ], tot_loss[loss=0.2995, simple_loss=0.3569, pruned_loss=0.1211, over 1609339.35 frames. ], batch size: 25, lr: 1.71e-02, grad_scale: 8.0 +2023-02-06 00:08:12,222 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8184, 2.1485, 1.6695, 2.7283, 1.3394, 1.3940, 1.6392, 2.2235], + device='cuda:0'), covar=tensor([0.1022, 0.1059, 0.1708, 0.0550, 0.1544, 0.2277, 0.1571, 0.1066], + device='cuda:0'), in_proj_covar=tensor([0.0272, 0.0275, 0.0298, 0.0224, 0.0260, 0.0286, 0.0297, 0.0268], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0004, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 00:08:26,147 INFO [optim.py:369] (0/4) 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 00:08:31,494 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31933.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:08:43,636 INFO [train.py:901] (0/4) Epoch 4, batch 7700, loss[loss=0.2545, simple_loss=0.3209, pruned_loss=0.0941, over 8238.00 frames. ], tot_loss[loss=0.3003, simple_loss=0.3574, pruned_loss=0.1216, over 1609162.55 frames. ], batch size: 22, lr: 1.71e-02, grad_scale: 8.0 +2023-02-06 00:08:43,815 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8910, 1.3675, 4.3508, 1.8095, 2.3919, 4.9155, 4.7682, 4.3951], + device='cuda:0'), covar=tensor([0.1239, 0.1529, 0.0238, 0.1807, 0.0757, 0.0211, 0.0306, 0.0465], + device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0271, 0.0225, 0.0268, 0.0224, 0.0202, 0.0214, 0.0274], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:0') +2023-02-06 00:08:51,019 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 +2023-02-06 00:08:55,989 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31968.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:08:56,026 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31968.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 00:09:06,798 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-06 00:09:10,169 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1945, 1.5573, 1.6101, 1.2533, 1.4522, 1.5347, 1.9865, 1.6940], + device='cuda:0'), covar=tensor([0.0675, 0.1337, 0.1857, 0.1569, 0.0665, 0.1539, 0.0782, 0.0612], + device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0185, 0.0228, 0.0191, 0.0142, 0.0193, 0.0155, 0.0159], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0007, 0.0008, 0.0007, 0.0005, 0.0007, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 00:09:17,589 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-32000.pt +2023-02-06 00:09:18,512 INFO [train.py:901] (0/4) Epoch 4, batch 7750, loss[loss=0.3394, simple_loss=0.4039, pruned_loss=0.1374, over 8293.00 frames. ], tot_loss[loss=0.3013, simple_loss=0.3583, pruned_loss=0.1222, over 1608694.05 frames. ], batch size: 23, lr: 1.71e-02, grad_scale: 8.0 +2023-02-06 00:09:35,964 INFO [optim.py:369] (0/4) 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 00:09:53,600 INFO [train.py:901] (0/4) Epoch 4, batch 7800, loss[loss=0.2989, simple_loss=0.3561, pruned_loss=0.1208, over 8249.00 frames. ], tot_loss[loss=0.3022, simple_loss=0.3586, pruned_loss=0.123, over 1605021.79 frames. ], batch size: 22, lr: 1.71e-02, grad_scale: 8.0 +2023-02-06 00:10:05,162 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=32066.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:10:27,386 INFO [train.py:901] (0/4) Epoch 4, batch 7850, loss[loss=0.2995, simple_loss=0.3409, pruned_loss=0.129, over 6354.00 frames. ], tot_loss[loss=0.3009, simple_loss=0.3575, pruned_loss=0.1221, over 1602723.11 frames. ], batch size: 14, lr: 1.71e-02, grad_scale: 8.0 +2023-02-06 00:10:43,942 INFO [optim.py:369] (0/4) 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 00:10:55,637 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=32142.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:11:00,792 INFO [train.py:901] (0/4) Epoch 4, batch 7900, loss[loss=0.3711, simple_loss=0.3858, pruned_loss=0.1782, over 7276.00 frames. ], tot_loss[loss=0.3024, simple_loss=0.3583, pruned_loss=0.1233, over 1601882.01 frames. ], batch size: 16, lr: 1.71e-02, grad_scale: 8.0 +2023-02-06 00:11:00,856 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=32150.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:11:12,554 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=32167.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:11:14,001 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.90 vs. limit=5.0 +2023-02-06 00:11:21,557 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32181.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:11:24,168 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=32185.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:11:34,457 INFO [train.py:901] (0/4) Epoch 4, batch 7950, loss[loss=0.2956, simple_loss=0.3508, pruned_loss=0.1202, over 8528.00 frames. ], tot_loss[loss=0.3049, simple_loss=0.3603, pruned_loss=0.1247, over 1605015.58 frames. ], batch size: 28, lr: 1.71e-02, grad_scale: 8.0 +2023-02-06 00:11:51,250 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=32224.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 00:11:51,630 INFO [optim.py:369] (0/4) 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 00:11:59,171 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.7603, 1.9479, 5.8694, 1.9557, 5.1158, 4.8650, 5.4010, 5.3730], + device='cuda:0'), covar=tensor([0.0335, 0.3257, 0.0214, 0.2195, 0.0779, 0.0529, 0.0296, 0.0349], + device='cuda:0'), in_proj_covar=tensor([0.0295, 0.0466, 0.0377, 0.0384, 0.0447, 0.0375, 0.0367, 0.0411], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 00:12:01,783 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=32240.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:12:08,285 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=32249.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 00:12:08,757 INFO [train.py:901] (0/4) Epoch 4, batch 8000, loss[loss=0.2556, simple_loss=0.3283, pruned_loss=0.09139, over 8744.00 frames. ], tot_loss[loss=0.3037, simple_loss=0.36, pruned_loss=0.1237, over 1606347.02 frames. ], batch size: 30, lr: 1.71e-02, grad_scale: 8.0 +2023-02-06 00:12:19,080 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32265.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:12:26,788 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=32277.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:12:42,459 INFO [train.py:901] (0/4) Epoch 4, batch 8050, loss[loss=0.2462, simple_loss=0.3065, pruned_loss=0.09295, over 7165.00 frames. ], tot_loss[loss=0.304, simple_loss=0.3592, pruned_loss=0.1244, over 1590082.76 frames. ], batch size: 16, lr: 1.70e-02, grad_scale: 8.0 +2023-02-06 00:12:42,636 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32300.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:12:42,668 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7340, 1.7908, 2.0293, 1.6231, 1.0636, 1.8402, 0.2974, 1.2260], + device='cuda:0'), covar=tensor([0.3833, 0.3243, 0.1391, 0.2300, 0.8030, 0.1833, 0.7732, 0.2743], + device='cuda:0'), in_proj_covar=tensor([0.0122, 0.0109, 0.0078, 0.0161, 0.0200, 0.0081, 0.0145, 0.0118], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 00:12:50,703 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=32312.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:12:52,791 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.9626, 2.6623, 3.9635, 3.1077, 3.1709, 2.5424, 1.6079, 1.9799], + device='cuda:0'), covar=tensor([0.1333, 0.1619, 0.0352, 0.0839, 0.0843, 0.0752, 0.0911, 0.1691], + device='cuda:0'), in_proj_covar=tensor([0.0695, 0.0627, 0.0521, 0.0595, 0.0707, 0.0581, 0.0567, 0.0579], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 00:12:58,915 INFO [optim.py:369] (0/4) 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 00:13:01,780 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3931, 1.9107, 3.0141, 2.3918, 2.3779, 1.8851, 1.3897, 1.0392], + device='cuda:0'), covar=tensor([0.1517, 0.1691, 0.0369, 0.0782, 0.0796, 0.0948, 0.1061, 0.1695], + device='cuda:0'), in_proj_covar=tensor([0.0700, 0.0632, 0.0526, 0.0596, 0.0710, 0.0585, 0.0570, 0.0583], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 00:13:04,772 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/epoch-4.pt +2023-02-06 00:13:16,071 WARNING [train.py:1067] (0/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-06 00:13:19,657 INFO [train.py:901] (0/4) Epoch 5, batch 0, loss[loss=0.4153, simple_loss=0.4495, pruned_loss=0.1905, over 8458.00 frames. ], tot_loss[loss=0.4153, simple_loss=0.4495, pruned_loss=0.1905, over 8458.00 frames. ], batch size: 27, lr: 1.59e-02, grad_scale: 8.0 +2023-02-06 00:13:19,658 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 00:13:31,621 INFO [train.py:935] (0/4) Epoch 5, validation: loss=0.2309, simple_loss=0.3254, pruned_loss=0.06822, over 944034.00 frames. +2023-02-06 00:13:31,622 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6689MB +2023-02-06 00:13:46,425 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-06 00:13:46,594 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32355.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:13:53,860 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.57 vs. limit=5.0 +2023-02-06 00:14:06,998 INFO [train.py:901] (0/4) Epoch 5, batch 50, loss[loss=0.3059, simple_loss=0.3523, pruned_loss=0.1297, over 8241.00 frames. ], tot_loss[loss=0.2982, simple_loss=0.3563, pruned_loss=0.1201, over 365399.09 frames. ], batch size: 22, lr: 1.58e-02, grad_scale: 8.0 +2023-02-06 00:14:13,361 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32392.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:14:20,370 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6540, 1.9924, 3.1353, 1.1934, 2.3018, 1.8863, 1.6972, 1.7335], + device='cuda:0'), covar=tensor([0.1340, 0.1640, 0.0609, 0.2953, 0.1271, 0.2270, 0.1316, 0.2203], + device='cuda:0'), in_proj_covar=tensor([0.0466, 0.0440, 0.0524, 0.0532, 0.0576, 0.0508, 0.0446, 0.0583], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:0') +2023-02-06 00:14:22,903 WARNING [train.py:1067] (0/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-06 00:14:36,522 INFO [optim.py:369] (0/4) 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 00:14:38,037 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32427.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:14:41,778 INFO [train.py:901] (0/4) Epoch 5, batch 100, loss[loss=0.3938, simple_loss=0.4234, pruned_loss=0.1821, over 8660.00 frames. ], tot_loss[loss=0.3001, simple_loss=0.3576, pruned_loss=0.1213, over 638641.78 frames. ], batch size: 34, lr: 1.58e-02, grad_scale: 8.0 +2023-02-06 00:14:44,562 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=32437.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:14:45,031 WARNING [train.py:1067] (0/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-06 00:14:56,153 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5165, 1.2311, 2.6781, 1.1923, 2.0366, 2.9502, 2.8075, 2.5317], + device='cuda:0'), covar=tensor([0.1130, 0.1459, 0.0456, 0.1934, 0.0722, 0.0344, 0.0453, 0.0736], + device='cuda:0'), in_proj_covar=tensor([0.0226, 0.0262, 0.0218, 0.0258, 0.0222, 0.0194, 0.0209, 0.0266], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:0') +2023-02-06 00:15:02,113 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=32462.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:15:11,551 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 +2023-02-06 00:15:15,808 INFO [train.py:901] (0/4) Epoch 5, batch 150, loss[loss=0.2184, simple_loss=0.2833, pruned_loss=0.07677, over 7426.00 frames. ], tot_loss[loss=0.2998, simple_loss=0.3585, pruned_loss=0.1205, over 858566.01 frames. ], batch size: 17, lr: 1.58e-02, grad_scale: 8.0 +2023-02-06 00:15:43,036 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=32521.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:15:45,468 INFO [optim.py:369] (0/4) 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 00:15:50,800 INFO [train.py:901] (0/4) Epoch 5, batch 200, loss[loss=0.2902, simple_loss=0.3384, pruned_loss=0.121, over 7806.00 frames. ], tot_loss[loss=0.2963, simple_loss=0.3558, pruned_loss=0.1184, over 1026726.09 frames. ], batch size: 20, lr: 1.58e-02, grad_scale: 8.0 +2023-02-06 00:15:59,813 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=32546.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:16:01,170 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2677, 1.8752, 1.9148, 0.7981, 1.8921, 1.2869, 0.4482, 1.7097], + device='cuda:0'), covar=tensor([0.0188, 0.0092, 0.0072, 0.0177, 0.0133, 0.0318, 0.0262, 0.0078], + device='cuda:0'), in_proj_covar=tensor([0.0292, 0.0205, 0.0165, 0.0252, 0.0200, 0.0329, 0.0269, 0.0234], + device='cuda:0'), out_proj_covar=tensor([1.1103e-04, 7.7410e-05, 6.0617e-05, 9.2566e-05, 7.6672e-05, 1.3401e-04, + 1.0360e-04, 8.7436e-05], device='cuda:0') +2023-02-06 00:16:06,402 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=32556.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:16:23,707 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=32581.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:16:24,862 INFO [train.py:901] (0/4) Epoch 5, batch 250, loss[loss=0.3542, simple_loss=0.4055, pruned_loss=0.1515, over 8325.00 frames. ], tot_loss[loss=0.2965, simple_loss=0.3557, pruned_loss=0.1187, over 1159360.17 frames. ], batch size: 25, lr: 1.58e-02, grad_scale: 8.0 +2023-02-06 00:16:36,290 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-06 00:16:45,136 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=32611.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:16:46,300 WARNING [train.py:1067] (0/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-06 00:16:54,491 INFO [optim.py:369] (0/4) 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 00:17:00,682 INFO [train.py:901] (0/4) Epoch 5, batch 300, loss[loss=0.2698, simple_loss=0.3406, pruned_loss=0.09953, over 8353.00 frames. ], tot_loss[loss=0.2971, simple_loss=0.356, pruned_loss=0.1191, over 1262036.48 frames. ], batch size: 24, lr: 1.58e-02, grad_scale: 8.0 +2023-02-06 00:17:03,006 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=32636.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:17:10,880 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=32648.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:17:27,573 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 5, batch 350, loss[loss=0.2951, simple_loss=0.3531, pruned_loss=0.1185, over 8032.00 frames. ], tot_loss[loss=0.2986, simple_loss=0.357, pruned_loss=0.1201, over 1340403.59 frames. ], batch size: 22, lr: 1.58e-02, grad_scale: 8.0 +2023-02-06 00:17:34,409 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=32683.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:17:51,720 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=32708.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:18:02,407 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9913, 2.5408, 3.3233, 0.9400, 3.1330, 2.0839, 1.3615, 1.7331], + device='cuda:0'), covar=tensor([0.0282, 0.0104, 0.0068, 0.0257, 0.0113, 0.0261, 0.0325, 0.0164], + device='cuda:0'), in_proj_covar=tensor([0.0288, 0.0201, 0.0165, 0.0248, 0.0197, 0.0328, 0.0267, 0.0230], + device='cuda:0'), out_proj_covar=tensor([1.0905e-04, 7.5591e-05, 5.9840e-05, 9.0840e-05, 7.4895e-05, 1.3290e-04, + 1.0252e-04, 8.5651e-05], device='cuda:0') +2023-02-06 00:18:04,027 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 5, batch 400, loss[loss=0.3751, simple_loss=0.4078, pruned_loss=0.1712, over 8336.00 frames. ], tot_loss[loss=0.3008, simple_loss=0.359, pruned_loss=0.1213, over 1402671.96 frames. ], batch size: 26, lr: 1.58e-02, grad_scale: 8.0 +2023-02-06 00:18:19,554 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.1291, 1.1250, 3.2027, 1.0064, 2.7436, 2.6801, 2.8969, 2.7973], + device='cuda:0'), covar=tensor([0.0493, 0.3265, 0.0625, 0.2459, 0.1346, 0.0852, 0.0580, 0.0723], + device='cuda:0'), in_proj_covar=tensor([0.0303, 0.0475, 0.0386, 0.0395, 0.0460, 0.0380, 0.0380, 0.0420], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 00:18:43,766 INFO [train.py:901] (0/4) Epoch 5, batch 450, loss[loss=0.3098, simple_loss=0.3707, pruned_loss=0.1245, over 7981.00 frames. ], tot_loss[loss=0.2982, simple_loss=0.3566, pruned_loss=0.1199, over 1448320.50 frames. ], batch size: 21, lr: 1.58e-02, grad_scale: 8.0 +2023-02-06 00:19:00,937 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 +2023-02-06 00:19:02,061 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0783, 1.2015, 1.1659, 0.1503, 1.1716, 0.9464, 0.1156, 1.0764], + device='cuda:0'), covar=tensor([0.0142, 0.0098, 0.0082, 0.0195, 0.0102, 0.0320, 0.0249, 0.0097], + device='cuda:0'), in_proj_covar=tensor([0.0287, 0.0204, 0.0163, 0.0247, 0.0199, 0.0326, 0.0268, 0.0229], + device='cuda:0'), out_proj_covar=tensor([1.0851e-04, 7.6543e-05, 5.8974e-05, 9.0359e-05, 7.5521e-05, 1.3192e-04, + 1.0272e-04, 8.5108e-05], device='cuda:0') +2023-02-06 00:19:12,444 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 5, batch 500, loss[loss=0.3228, simple_loss=0.3749, pruned_loss=0.1353, over 8289.00 frames. ], tot_loss[loss=0.2971, simple_loss=0.3559, pruned_loss=0.1191, over 1484139.13 frames. ], batch size: 23, lr: 1.57e-02, grad_scale: 8.0 +2023-02-06 00:19:21,935 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4182, 1.7883, 1.7768, 1.5934, 0.8554, 1.8095, 0.2034, 1.2030], + device='cuda:0'), covar=tensor([0.4197, 0.2235, 0.1329, 0.2265, 0.8307, 0.1158, 0.5854, 0.2101], + device='cuda:0'), in_proj_covar=tensor([0.0125, 0.0112, 0.0079, 0.0162, 0.0208, 0.0078, 0.0143, 0.0118], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 00:19:52,903 INFO [train.py:901] (0/4) Epoch 5, batch 550, loss[loss=0.2543, simple_loss=0.3345, pruned_loss=0.08704, over 8506.00 frames. ], tot_loss[loss=0.2962, simple_loss=0.3553, pruned_loss=0.1185, over 1515478.39 frames. ], batch size: 26, lr: 1.57e-02, grad_scale: 8.0 +2023-02-06 00:19:55,207 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3838, 1.6757, 1.6021, 1.3526, 0.8573, 1.6143, 0.2008, 1.0529], + device='cuda:0'), covar=tensor([0.3693, 0.1466, 0.1275, 0.1946, 0.5978, 0.1037, 0.4862, 0.1923], + device='cuda:0'), in_proj_covar=tensor([0.0126, 0.0113, 0.0079, 0.0164, 0.0211, 0.0079, 0.0145, 0.0119], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 00:20:21,235 INFO [optim.py:369] (0/4) 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 00:20:23,468 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6416, 1.3990, 2.7297, 1.0319, 2.0610, 3.0365, 2.8424, 2.4805], + device='cuda:0'), covar=tensor([0.0938, 0.1325, 0.0487, 0.2177, 0.0700, 0.0298, 0.0508, 0.0745], + device='cuda:0'), in_proj_covar=tensor([0.0230, 0.0263, 0.0223, 0.0261, 0.0224, 0.0195, 0.0215, 0.0272], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:0') +2023-02-06 00:20:26,722 INFO [train.py:901] (0/4) Epoch 5, batch 600, loss[loss=0.3224, simple_loss=0.3778, pruned_loss=0.1335, over 8456.00 frames. ], tot_loss[loss=0.2946, simple_loss=0.3542, pruned_loss=0.1175, over 1537674.72 frames. ], batch size: 27, lr: 1.57e-02, grad_scale: 8.0 +2023-02-06 00:20:50,096 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-06 00:21:02,155 INFO [train.py:901] (0/4) Epoch 5, batch 650, loss[loss=0.3425, simple_loss=0.4029, pruned_loss=0.141, over 8476.00 frames. ], tot_loss[loss=0.2924, simple_loss=0.3529, pruned_loss=0.116, over 1556110.21 frames. ], batch size: 25, lr: 1.57e-02, grad_scale: 8.0 +2023-02-06 00:21:02,982 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32984.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:21:18,950 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6584, 1.5835, 3.4038, 1.1370, 2.2479, 3.9628, 3.5509, 3.3453], + device='cuda:0'), covar=tensor([0.1157, 0.1305, 0.0349, 0.2071, 0.0765, 0.0190, 0.0367, 0.0574], + device='cuda:0'), in_proj_covar=tensor([0.0230, 0.0261, 0.0221, 0.0260, 0.0225, 0.0193, 0.0212, 0.0270], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:0') +2023-02-06 00:21:28,368 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9747, 2.1947, 1.9770, 2.5951, 1.7902, 1.8552, 1.9860, 2.3821], + device='cuda:0'), covar=tensor([0.0983, 0.1141, 0.1237, 0.0670, 0.1338, 0.1548, 0.1245, 0.0937], + device='cuda:0'), in_proj_covar=tensor([0.0274, 0.0274, 0.0294, 0.0220, 0.0255, 0.0285, 0.0291, 0.0269], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0004, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 00:21:30,782 INFO [optim.py:369] (0/4) 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 00:21:35,148 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-02-06 00:21:36,137 INFO [train.py:901] (0/4) Epoch 5, batch 700, loss[loss=0.2894, simple_loss=0.357, pruned_loss=0.1108, over 8109.00 frames. ], tot_loss[loss=0.2928, simple_loss=0.3531, pruned_loss=0.1162, over 1566106.90 frames. ], batch size: 23, lr: 1.57e-02, grad_scale: 8.0 +2023-02-06 00:22:11,097 INFO [train.py:901] (0/4) Epoch 5, batch 750, loss[loss=0.3142, simple_loss=0.3806, pruned_loss=0.1239, over 8358.00 frames. ], tot_loss[loss=0.2948, simple_loss=0.3549, pruned_loss=0.1174, over 1577127.12 frames. ], batch size: 24, lr: 1.57e-02, grad_scale: 8.0 +2023-02-06 00:22:14,419 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33087.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:22:36,874 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-06 00:22:40,967 INFO [optim.py:369] (0/4) 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 00:22:45,500 WARNING [train.py:1067] (0/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-06 00:22:46,149 INFO [train.py:901] (0/4) Epoch 5, batch 800, loss[loss=0.3045, simple_loss=0.3544, pruned_loss=0.1273, over 8024.00 frames. ], tot_loss[loss=0.2947, simple_loss=0.3546, pruned_loss=0.1174, over 1585408.01 frames. ], batch size: 22, lr: 1.57e-02, grad_scale: 8.0 +2023-02-06 00:23:13,714 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 5, batch 850, loss[loss=0.3298, simple_loss=0.385, pruned_loss=0.1373, over 8144.00 frames. ], tot_loss[loss=0.2957, simple_loss=0.3551, pruned_loss=0.1182, over 1594443.75 frames. ], batch size: 22, lr: 1.57e-02, grad_scale: 8.0 +2023-02-06 00:23:49,971 INFO [optim.py:369] (0/4) 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 00:23:56,030 INFO [train.py:901] (0/4) Epoch 5, batch 900, loss[loss=0.2942, simple_loss=0.3618, pruned_loss=0.1133, over 8120.00 frames. ], tot_loss[loss=0.2939, simple_loss=0.3532, pruned_loss=0.1173, over 1594201.81 frames. ], batch size: 22, lr: 1.57e-02, grad_scale: 8.0 +2023-02-06 00:24:29,744 INFO [train.py:901] (0/4) Epoch 5, batch 950, loss[loss=0.3055, simple_loss=0.3676, pruned_loss=0.1217, over 8463.00 frames. ], tot_loss[loss=0.2938, simple_loss=0.3537, pruned_loss=0.117, over 1601273.49 frames. ], batch size: 25, lr: 1.56e-02, grad_scale: 8.0 +2023-02-06 00:25:01,022 INFO [optim.py:369] (0/4) 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 00:25:03,061 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=33328.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:25:05,099 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-06 00:25:06,446 INFO [train.py:901] (0/4) Epoch 5, batch 1000, loss[loss=0.2645, simple_loss=0.3336, pruned_loss=0.09771, over 8187.00 frames. ], tot_loss[loss=0.2923, simple_loss=0.3526, pruned_loss=0.116, over 1605136.59 frames. ], batch size: 23, lr: 1.56e-02, grad_scale: 8.0 +2023-02-06 00:25:34,765 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 +2023-02-06 00:25:40,441 INFO [train.py:901] (0/4) Epoch 5, batch 1050, loss[loss=0.3066, simple_loss=0.3678, pruned_loss=0.1227, over 8312.00 frames. ], tot_loss[loss=0.2919, simple_loss=0.3514, pruned_loss=0.1162, over 1599833.65 frames. ], batch size: 25, lr: 1.56e-02, grad_scale: 8.0 +2023-02-06 00:25:40,475 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-06 00:25:52,226 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-06 00:26:08,785 INFO [optim.py:369] (0/4) 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 00:26:13,582 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=33431.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:26:14,846 INFO [train.py:901] (0/4) Epoch 5, batch 1100, loss[loss=0.3108, simple_loss=0.3815, pruned_loss=0.1201, over 8626.00 frames. ], tot_loss[loss=0.2912, simple_loss=0.3511, pruned_loss=0.1157, over 1605711.58 frames. ], batch size: 49, lr: 1.56e-02, grad_scale: 8.0 +2023-02-06 00:26:22,995 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=33443.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:26:29,012 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7434, 1.9723, 2.1484, 1.7005, 1.1214, 2.1261, 0.2904, 1.1122], + device='cuda:0'), covar=tensor([0.2660, 0.2359, 0.0847, 0.2379, 0.7015, 0.0790, 0.6195, 0.2987], + device='cuda:0'), in_proj_covar=tensor([0.0124, 0.0116, 0.0079, 0.0162, 0.0208, 0.0080, 0.0144, 0.0121], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 00:26:50,043 INFO [train.py:901] (0/4) Epoch 5, batch 1150, loss[loss=0.2968, simple_loss=0.3659, pruned_loss=0.1139, over 8740.00 frames. ], tot_loss[loss=0.2917, simple_loss=0.3518, pruned_loss=0.1159, over 1610559.06 frames. ], batch size: 30, lr: 1.56e-02, grad_scale: 8.0 +2023-02-06 00:26:54,912 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33490.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:27:02,059 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-06 00:27:13,077 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=33517.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:27:18,265 INFO [optim.py:369] (0/4) 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 00:27:23,603 INFO [train.py:901] (0/4) Epoch 5, batch 1200, loss[loss=0.31, simple_loss=0.3543, pruned_loss=0.1329, over 8633.00 frames. ], tot_loss[loss=0.2918, simple_loss=0.3518, pruned_loss=0.1159, over 1607940.29 frames. ], batch size: 49, lr: 1.56e-02, grad_scale: 8.0 +2023-02-06 00:27:32,543 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=33546.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:27:53,514 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.5417, 5.5118, 4.9208, 1.6677, 4.8321, 5.1883, 5.2531, 4.6131], + device='cuda:0'), covar=tensor([0.0724, 0.0550, 0.0893, 0.5394, 0.0699, 0.0909, 0.1200, 0.0964], + device='cuda:0'), in_proj_covar=tensor([0.0375, 0.0281, 0.0303, 0.0393, 0.0301, 0.0262, 0.0286, 0.0232], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:0') +2023-02-06 00:28:00,128 INFO [train.py:901] (0/4) Epoch 5, batch 1250, loss[loss=0.2388, simple_loss=0.3073, pruned_loss=0.08514, over 8083.00 frames. ], tot_loss[loss=0.2931, simple_loss=0.3527, pruned_loss=0.1167, over 1610974.76 frames. ], batch size: 21, lr: 1.56e-02, grad_scale: 8.0 +2023-02-06 00:28:15,585 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.8119, 3.7305, 3.4057, 1.6806, 3.3242, 3.3137, 3.4436, 2.8755], + device='cuda:0'), covar=tensor([0.0941, 0.0719, 0.0997, 0.4265, 0.0809, 0.0935, 0.1316, 0.0948], + device='cuda:0'), in_proj_covar=tensor([0.0371, 0.0277, 0.0300, 0.0390, 0.0296, 0.0260, 0.0287, 0.0231], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:0') +2023-02-06 00:28:29,031 INFO [optim.py:369] (0/4) 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 00:28:34,050 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=33632.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:28:34,562 INFO [train.py:901] (0/4) Epoch 5, batch 1300, loss[loss=0.268, simple_loss=0.3267, pruned_loss=0.1047, over 8092.00 frames. ], tot_loss[loss=0.2937, simple_loss=0.3532, pruned_loss=0.1172, over 1608499.17 frames. ], batch size: 21, lr: 1.56e-02, grad_scale: 8.0 +2023-02-06 00:29:10,156 INFO [train.py:901] (0/4) Epoch 5, batch 1350, loss[loss=0.346, simple_loss=0.3938, pruned_loss=0.1491, over 8295.00 frames. ], tot_loss[loss=0.2965, simple_loss=0.3554, pruned_loss=0.1188, over 1610390.12 frames. ], batch size: 23, lr: 1.55e-02, grad_scale: 4.0 +2023-02-06 00:29:18,147 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33695.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:29:21,565 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33699.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:29:30,025 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 +2023-02-06 00:29:38,302 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33724.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:29:39,358 INFO [optim.py:369] (0/4) 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 00:29:44,201 INFO [train.py:901] (0/4) Epoch 5, batch 1400, loss[loss=0.2868, simple_loss=0.3561, pruned_loss=0.1087, over 8456.00 frames. ], tot_loss[loss=0.2961, simple_loss=0.3555, pruned_loss=0.1183, over 1609557.40 frames. ], batch size: 25, lr: 1.55e-02, grad_scale: 4.0 +2023-02-06 00:29:52,408 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.7776, 2.5588, 2.9134, 2.3438, 1.5031, 2.8243, 0.7253, 1.7397], + device='cuda:0'), covar=tensor([0.3008, 0.2926, 0.0892, 0.2181, 0.6134, 0.0808, 0.5784, 0.2377], + device='cuda:0'), in_proj_covar=tensor([0.0129, 0.0117, 0.0081, 0.0167, 0.0212, 0.0081, 0.0148, 0.0125], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 00:30:18,167 INFO [train.py:901] (0/4) Epoch 5, batch 1450, loss[loss=0.3151, simple_loss=0.3775, pruned_loss=0.1264, over 8515.00 frames. ], tot_loss[loss=0.297, simple_loss=0.3566, pruned_loss=0.1187, over 1614651.82 frames. ], batch size: 26, lr: 1.55e-02, grad_scale: 4.0 +2023-02-06 00:30:32,288 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-06 00:30:32,501 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33802.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:30:49,119 INFO [optim.py:369] (0/4) 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 00:30:50,045 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33827.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:30:54,036 INFO [train.py:901] (0/4) Epoch 5, batch 1500, loss[loss=0.3347, simple_loss=0.3854, pruned_loss=0.142, over 7820.00 frames. ], tot_loss[loss=0.2944, simple_loss=0.3545, pruned_loss=0.1172, over 1614103.97 frames. ], batch size: 20, lr: 1.55e-02, grad_scale: 4.0 +2023-02-06 00:30:54,786 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=33834.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:31:09,215 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-02-06 00:31:27,613 INFO [train.py:901] (0/4) Epoch 5, batch 1550, loss[loss=0.2238, simple_loss=0.3003, pruned_loss=0.07369, over 8083.00 frames. ], tot_loss[loss=0.2955, simple_loss=0.3555, pruned_loss=0.1178, over 1619801.46 frames. ], batch size: 21, lr: 1.55e-02, grad_scale: 4.0 +2023-02-06 00:31:31,093 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33888.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:31:48,411 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33913.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:31:58,126 INFO [optim.py:369] (0/4) 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 00:32:03,152 INFO [train.py:901] (0/4) Epoch 5, batch 1600, loss[loss=0.2891, simple_loss=0.3609, pruned_loss=0.1087, over 8291.00 frames. ], tot_loss[loss=0.2961, simple_loss=0.3557, pruned_loss=0.1183, over 1619440.90 frames. ], batch size: 23, lr: 1.55e-02, grad_scale: 8.0 +2023-02-06 00:32:12,809 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6945, 1.4344, 2.9812, 1.3401, 2.2543, 3.2168, 3.1325, 2.8110], + device='cuda:0'), covar=tensor([0.1086, 0.1387, 0.0431, 0.1988, 0.0695, 0.0308, 0.0413, 0.0631], + device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0266, 0.0216, 0.0260, 0.0221, 0.0195, 0.0220, 0.0270], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:0') +2023-02-06 00:32:14,803 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=33949.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:32:20,115 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33957.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:32:30,687 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8965, 1.4698, 3.4234, 1.4746, 2.3191, 3.8362, 3.6120, 3.2998], + device='cuda:0'), covar=tensor([0.1058, 0.1498, 0.0307, 0.1862, 0.0790, 0.0284, 0.0354, 0.0551], + device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0266, 0.0217, 0.0259, 0.0222, 0.0194, 0.0220, 0.0272], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:0') +2023-02-06 00:32:37,058 INFO [train.py:901] (0/4) Epoch 5, batch 1650, loss[loss=0.274, simple_loss=0.3245, pruned_loss=0.1118, over 7425.00 frames. ], tot_loss[loss=0.2926, simple_loss=0.3528, pruned_loss=0.1162, over 1618541.71 frames. ], batch size: 17, lr: 1.55e-02, grad_scale: 8.0 +2023-02-06 00:32:48,559 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-34000.pt +2023-02-06 00:33:07,304 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 5, batch 1700, loss[loss=0.2791, simple_loss=0.3539, pruned_loss=0.1022, over 8499.00 frames. ], tot_loss[loss=0.293, simple_loss=0.353, pruned_loss=0.1165, over 1616792.15 frames. ], batch size: 26, lr: 1.55e-02, grad_scale: 8.0 +2023-02-06 00:33:17,258 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=34039.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:33:33,629 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34062.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:33:39,804 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.7136, 5.6482, 4.9920, 2.2702, 5.0965, 5.3082, 5.3030, 4.8643], + device='cuda:0'), covar=tensor([0.0610, 0.0486, 0.0716, 0.4263, 0.0543, 0.0534, 0.1057, 0.0471], + device='cuda:0'), in_proj_covar=tensor([0.0392, 0.0294, 0.0311, 0.0405, 0.0313, 0.0273, 0.0302, 0.0244], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:0') +2023-02-06 00:33:44,837 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 +2023-02-06 00:33:47,846 INFO [train.py:901] (0/4) Epoch 5, batch 1750, loss[loss=0.2853, simple_loss=0.3507, pruned_loss=0.11, over 8352.00 frames. ], tot_loss[loss=0.2923, simple_loss=0.3527, pruned_loss=0.116, over 1621559.54 frames. ], batch size: 24, lr: 1.55e-02, grad_scale: 8.0 +2023-02-06 00:34:16,516 INFO [optim.py:369] (0/4) 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 00:34:21,845 INFO [train.py:901] (0/4) Epoch 5, batch 1800, loss[loss=0.2978, simple_loss=0.366, pruned_loss=0.1148, over 8290.00 frames. ], tot_loss[loss=0.292, simple_loss=0.3522, pruned_loss=0.1159, over 1615987.31 frames. ], batch size: 23, lr: 1.54e-02, grad_scale: 8.0 +2023-02-06 00:34:24,330 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-02-06 00:34:36,703 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34154.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:34:43,245 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34163.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:34:57,289 INFO [train.py:901] (0/4) Epoch 5, batch 1850, loss[loss=0.2516, simple_loss=0.3317, pruned_loss=0.08574, over 8480.00 frames. ], tot_loss[loss=0.2921, simple_loss=0.3526, pruned_loss=0.1158, over 1617669.14 frames. ], batch size: 27, lr: 1.54e-02, grad_scale: 8.0 +2023-02-06 00:35:12,354 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34205.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:35:26,207 INFO [optim.py:369] (0/4) 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 00:35:29,068 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34230.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:35:30,895 INFO [train.py:901] (0/4) Epoch 5, batch 1900, loss[loss=0.2885, simple_loss=0.3584, pruned_loss=0.1093, over 8342.00 frames. ], tot_loss[loss=0.292, simple_loss=0.3526, pruned_loss=0.1157, over 1617969.96 frames. ], batch size: 26, lr: 1.54e-02, grad_scale: 8.0 +2023-02-06 00:36:00,402 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 +2023-02-06 00:36:05,944 INFO [train.py:901] (0/4) Epoch 5, batch 1950, loss[loss=0.3154, simple_loss=0.3721, pruned_loss=0.1293, over 8029.00 frames. ], tot_loss[loss=0.2932, simple_loss=0.3535, pruned_loss=0.1165, over 1616680.35 frames. ], batch size: 22, lr: 1.54e-02, grad_scale: 8.0 +2023-02-06 00:36:09,866 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-06 00:36:18,691 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=34301.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:36:23,389 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-06 00:36:35,385 INFO [optim.py:369] (0/4) 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 00:36:40,164 INFO [train.py:901] (0/4) Epoch 5, batch 2000, loss[loss=0.2833, simple_loss=0.3446, pruned_loss=0.111, over 8103.00 frames. ], tot_loss[loss=0.2926, simple_loss=0.353, pruned_loss=0.1161, over 1615777.45 frames. ], batch size: 23, lr: 1.54e-02, grad_scale: 8.0 +2023-02-06 00:36:42,291 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 5, batch 2050, loss[loss=0.2918, simple_loss=0.3385, pruned_loss=0.1226, over 7695.00 frames. ], tot_loss[loss=0.2945, simple_loss=0.3544, pruned_loss=0.1173, over 1612140.44 frames. ], batch size: 18, lr: 1.54e-02, grad_scale: 8.0 +2023-02-06 00:37:31,116 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=34406.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:37:33,923 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34410.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:37:38,675 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) Epoch 5, batch 2100, loss[loss=0.2668, simple_loss=0.3449, pruned_loss=0.09438, over 8632.00 frames. ], tot_loss[loss=0.2948, simple_loss=0.3542, pruned_loss=0.1178, over 1610322.14 frames. ], batch size: 34, lr: 1.54e-02, grad_scale: 8.0 +2023-02-06 00:37:51,412 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34435.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:37:55,819 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 +2023-02-06 00:38:05,709 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 +2023-02-06 00:38:23,172 INFO [train.py:901] (0/4) Epoch 5, batch 2150, loss[loss=0.3113, simple_loss=0.381, pruned_loss=0.1207, over 8314.00 frames. ], tot_loss[loss=0.2923, simple_loss=0.3521, pruned_loss=0.1163, over 1610359.80 frames. ], batch size: 26, lr: 1.54e-02, grad_scale: 8.0 +2023-02-06 00:38:30,787 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3184, 1.9555, 3.0697, 2.6226, 2.5367, 1.9607, 1.3853, 1.1021], + device='cuda:0'), covar=tensor([0.1554, 0.1760, 0.0394, 0.0785, 0.0717, 0.0901, 0.0961, 0.1797], + device='cuda:0'), in_proj_covar=tensor([0.0723, 0.0660, 0.0563, 0.0643, 0.0749, 0.0614, 0.0590, 0.0618], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 00:38:39,929 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=34507.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:38:46,882 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 +2023-02-06 00:38:50,630 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) Epoch 5, batch 2200, loss[loss=0.307, simple_loss=0.3545, pruned_loss=0.1297, over 8235.00 frames. ], tot_loss[loss=0.2905, simple_loss=0.3508, pruned_loss=0.1151, over 1611631.80 frames. ], batch size: 22, lr: 1.54e-02, grad_scale: 8.0 +2023-02-06 00:39:16,174 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6474, 3.7949, 2.0066, 2.4012, 2.6449, 1.6385, 2.2327, 2.8971], + device='cuda:0'), covar=tensor([0.1461, 0.0249, 0.1010, 0.0773, 0.0649, 0.1146, 0.1019, 0.0815], + device='cuda:0'), in_proj_covar=tensor([0.0363, 0.0253, 0.0323, 0.0315, 0.0326, 0.0314, 0.0346, 0.0317], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 00:39:32,422 INFO [train.py:901] (0/4) Epoch 5, batch 2250, loss[loss=0.2695, simple_loss=0.349, pruned_loss=0.09494, over 8448.00 frames. ], tot_loss[loss=0.2919, simple_loss=0.3516, pruned_loss=0.1161, over 1609791.05 frames. ], batch size: 27, lr: 1.53e-02, grad_scale: 8.0 +2023-02-06 00:39:37,138 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.1990, 1.0532, 3.2728, 0.9272, 2.8626, 2.7452, 2.9084, 2.8265], + device='cuda:0'), covar=tensor([0.0475, 0.3287, 0.0513, 0.2559, 0.1185, 0.0715, 0.0577, 0.0698], + device='cuda:0'), in_proj_covar=tensor([0.0323, 0.0488, 0.0399, 0.0410, 0.0481, 0.0397, 0.0400, 0.0441], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 00:39:52,216 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34611.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:39:59,486 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34622.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:40:02,640 INFO [optim.py:369] (0/4) 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 00:40:07,917 INFO [train.py:901] (0/4) Epoch 5, batch 2300, loss[loss=0.3418, simple_loss=0.3891, pruned_loss=0.1473, over 8344.00 frames. ], tot_loss[loss=0.2941, simple_loss=0.3535, pruned_loss=0.1174, over 1612962.01 frames. ], batch size: 26, lr: 1.53e-02, grad_scale: 8.0 +2023-02-06 00:40:12,715 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34640.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:40:16,079 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34644.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:40:34,977 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34672.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:40:42,256 INFO [train.py:901] (0/4) Epoch 5, batch 2350, loss[loss=0.2899, simple_loss=0.3599, pruned_loss=0.11, over 8518.00 frames. ], tot_loss[loss=0.2945, simple_loss=0.3541, pruned_loss=0.1175, over 1614244.30 frames. ], batch size: 49, lr: 1.53e-02, grad_scale: 8.0 +2023-02-06 00:40:51,694 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34697.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:41:11,438 INFO [optim.py:369] (0/4) 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 00:41:12,928 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34728.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 00:41:16,110 INFO [train.py:901] (0/4) Epoch 5, batch 2400, loss[loss=0.3336, simple_loss=0.3913, pruned_loss=0.138, over 8572.00 frames. ], tot_loss[loss=0.2943, simple_loss=0.3533, pruned_loss=0.1176, over 1610855.19 frames. ], batch size: 31, lr: 1.53e-02, grad_scale: 8.0 +2023-02-06 00:41:47,413 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34777.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:41:51,214 INFO [train.py:901] (0/4) Epoch 5, batch 2450, loss[loss=0.2915, simple_loss=0.3427, pruned_loss=0.1201, over 8047.00 frames. ], tot_loss[loss=0.2942, simple_loss=0.3537, pruned_loss=0.1174, over 1613821.26 frames. ], batch size: 20, lr: 1.53e-02, grad_scale: 8.0 +2023-02-06 00:41:51,353 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.7122, 1.4497, 5.7844, 2.1956, 5.0334, 4.8367, 5.2238, 5.1915], + device='cuda:0'), covar=tensor([0.0371, 0.3848, 0.0222, 0.2263, 0.0885, 0.0492, 0.0417, 0.0438], + device='cuda:0'), in_proj_covar=tensor([0.0316, 0.0488, 0.0400, 0.0406, 0.0474, 0.0395, 0.0396, 0.0436], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 00:41:55,099 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 +2023-02-06 00:42:04,191 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34802.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:42:19,453 INFO [optim.py:369] (0/4) 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 00:42:24,144 INFO [train.py:901] (0/4) Epoch 5, batch 2500, loss[loss=0.1958, simple_loss=0.2628, pruned_loss=0.06436, over 7694.00 frames. ], tot_loss[loss=0.2934, simple_loss=0.3527, pruned_loss=0.117, over 1612122.88 frames. ], batch size: 18, lr: 1.53e-02, grad_scale: 8.0 +2023-02-06 00:42:43,923 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0784, 1.1301, 1.1030, 0.9011, 0.7296, 1.1468, 0.0160, 0.9630], + device='cuda:0'), covar=tensor([0.3352, 0.2381, 0.1421, 0.2476, 0.6424, 0.1113, 0.5518, 0.2263], + device='cuda:0'), in_proj_covar=tensor([0.0126, 0.0123, 0.0080, 0.0164, 0.0207, 0.0081, 0.0147, 0.0120], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 00:42:55,972 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 5, batch 2550, loss[loss=0.2348, simple_loss=0.3088, pruned_loss=0.08041, over 8234.00 frames. ], tot_loss[loss=0.2946, simple_loss=0.3536, pruned_loss=0.1178, over 1611398.45 frames. ], batch size: 22, lr: 1.53e-02, grad_scale: 8.0 +2023-02-06 00:43:13,472 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34903.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:43:29,141 INFO [optim.py:369] (0/4) 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 00:43:33,913 INFO [train.py:901] (0/4) Epoch 5, batch 2600, loss[loss=0.3123, simple_loss=0.3718, pruned_loss=0.1264, over 7803.00 frames. ], tot_loss[loss=0.2971, simple_loss=0.3555, pruned_loss=0.1193, over 1612276.61 frames. ], batch size: 20, lr: 1.53e-02, grad_scale: 8.0 +2023-02-06 00:43:49,004 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=34955.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:44:00,259 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.5460, 3.5575, 3.1771, 2.0397, 3.2128, 3.1154, 3.2867, 2.7988], + device='cuda:0'), covar=tensor([0.0909, 0.0680, 0.0950, 0.3514, 0.0733, 0.0829, 0.1186, 0.0804], + device='cuda:0'), in_proj_covar=tensor([0.0391, 0.0281, 0.0310, 0.0400, 0.0304, 0.0269, 0.0292, 0.0237], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:0') +2023-02-06 00:44:09,533 INFO [train.py:901] (0/4) Epoch 5, batch 2650, loss[loss=0.3211, simple_loss=0.3834, pruned_loss=0.1294, over 8544.00 frames. ], tot_loss[loss=0.2954, simple_loss=0.3544, pruned_loss=0.1182, over 1611975.68 frames. ], batch size: 39, lr: 1.53e-02, grad_scale: 8.0 +2023-02-06 00:44:10,272 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=34984.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:44:13,017 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=34988.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:44:33,030 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 +2023-02-06 00:44:39,302 INFO [optim.py:369] (0/4) 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 00:44:43,764 INFO [train.py:901] (0/4) Epoch 5, batch 2700, loss[loss=0.3251, simple_loss=0.3811, pruned_loss=0.1345, over 8125.00 frames. ], tot_loss[loss=0.2945, simple_loss=0.3538, pruned_loss=0.1176, over 1612460.22 frames. ], batch size: 22, lr: 1.53e-02, grad_scale: 8.0 +2023-02-06 00:45:09,474 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35070.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:45:10,729 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=35072.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 00:45:18,332 INFO [train.py:901] (0/4) Epoch 5, batch 2750, loss[loss=0.347, simple_loss=0.3993, pruned_loss=0.1474, over 8491.00 frames. ], tot_loss[loss=0.2934, simple_loss=0.353, pruned_loss=0.1168, over 1613582.54 frames. ], batch size: 28, lr: 1.52e-02, grad_scale: 8.0 +2023-02-06 00:45:29,743 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35099.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:45:32,506 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35103.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:45:37,935 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35110.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:45:48,687 INFO [optim.py:369] (0/4) 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 00:45:53,667 INFO [train.py:901] (0/4) Epoch 5, batch 2800, loss[loss=0.3071, simple_loss=0.36, pruned_loss=0.1271, over 7798.00 frames. ], tot_loss[loss=0.295, simple_loss=0.3549, pruned_loss=0.1176, over 1618443.51 frames. ], batch size: 20, lr: 1.52e-02, grad_scale: 8.0 +2023-02-06 00:46:27,512 INFO [train.py:901] (0/4) Epoch 5, batch 2850, loss[loss=0.3283, simple_loss=0.3766, pruned_loss=0.14, over 8649.00 frames. ], tot_loss[loss=0.2958, simple_loss=0.3551, pruned_loss=0.1183, over 1620896.80 frames. ], batch size: 39, lr: 1.52e-02, grad_scale: 8.0 +2023-02-06 00:46:30,504 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35187.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 00:46:58,368 INFO [optim.py:369] (0/4) 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 00:47:03,662 INFO [train.py:901] (0/4) Epoch 5, batch 2900, loss[loss=0.3421, simple_loss=0.3809, pruned_loss=0.1516, over 7916.00 frames. ], tot_loss[loss=0.2969, simple_loss=0.3562, pruned_loss=0.1188, over 1621784.31 frames. ], batch size: 20, lr: 1.52e-02, grad_scale: 8.0 +2023-02-06 00:47:16,753 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.05 vs. limit=5.0 +2023-02-06 00:47:36,540 INFO [train.py:901] (0/4) Epoch 5, batch 2950, loss[loss=0.3347, simple_loss=0.3926, pruned_loss=0.1383, over 8570.00 frames. ], tot_loss[loss=0.2956, simple_loss=0.3554, pruned_loss=0.1179, over 1622430.94 frames. ], batch size: 39, lr: 1.52e-02, grad_scale: 8.0 +2023-02-06 00:47:41,876 WARNING [train.py:1067] (0/4) 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] (0/4) 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 00:48:07,075 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35326.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:48:12,107 INFO [train.py:901] (0/4) Epoch 5, batch 3000, loss[loss=0.2249, simple_loss=0.2931, pruned_loss=0.07832, over 7648.00 frames. ], tot_loss[loss=0.2962, simple_loss=0.3556, pruned_loss=0.1184, over 1621384.04 frames. ], batch size: 19, lr: 1.52e-02, grad_scale: 8.0 +2023-02-06 00:48:12,108 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 00:48:25,509 INFO [train.py:935] (0/4) Epoch 5, validation: loss=0.2228, simple_loss=0.319, pruned_loss=0.0633, over 944034.00 frames. +2023-02-06 00:48:25,510 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6689MB +2023-02-06 00:48:39,249 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35351.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:48:41,982 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35355.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:48:44,687 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35359.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:48:47,325 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35363.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 00:48:59,229 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35380.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:49:01,090 INFO [train.py:901] (0/4) Epoch 5, batch 3050, loss[loss=0.2717, simple_loss=0.3412, pruned_loss=0.1011, over 7964.00 frames. ], tot_loss[loss=0.2955, simple_loss=0.3552, pruned_loss=0.1179, over 1620152.36 frames. ], batch size: 21, lr: 1.52e-02, grad_scale: 8.0 +2023-02-06 00:49:01,941 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35384.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:49:07,184 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35392.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:49:20,088 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5897, 2.2652, 4.5210, 1.1839, 2.8129, 2.0281, 1.5625, 2.4734], + device='cuda:0'), covar=tensor([0.1573, 0.1778, 0.0696, 0.3123, 0.1470, 0.2469, 0.1466, 0.2453], + device='cuda:0'), in_proj_covar=tensor([0.0473, 0.0448, 0.0531, 0.0535, 0.0577, 0.0515, 0.0455, 0.0589], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:0') +2023-02-06 00:49:29,664 INFO [optim.py:369] (0/4) 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 00:49:34,188 INFO [train.py:901] (0/4) Epoch 5, batch 3100, loss[loss=0.2633, simple_loss=0.3239, pruned_loss=0.1013, over 7799.00 frames. ], tot_loss[loss=0.2941, simple_loss=0.354, pruned_loss=0.1171, over 1619599.19 frames. ], batch size: 19, lr: 1.52e-02, grad_scale: 8.0 +2023-02-06 00:49:41,030 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35443.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 00:49:48,035 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=35454.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:49:59,629 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35468.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 00:50:09,986 INFO [train.py:901] (0/4) Epoch 5, batch 3150, loss[loss=0.2885, simple_loss=0.3604, pruned_loss=0.1083, over 8503.00 frames. ], tot_loss[loss=0.2936, simple_loss=0.354, pruned_loss=0.1166, over 1620899.80 frames. ], batch size: 26, lr: 1.52e-02, grad_scale: 8.0 +2023-02-06 00:50:10,784 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.0414, 2.8307, 2.7009, 1.4816, 2.7357, 2.6685, 2.7976, 2.4106], + device='cuda:0'), covar=tensor([0.1176, 0.0879, 0.1091, 0.4552, 0.0988, 0.1212, 0.1293, 0.1132], + device='cuda:0'), in_proj_covar=tensor([0.0392, 0.0283, 0.0318, 0.0407, 0.0310, 0.0273, 0.0297, 0.0242], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:0') +2023-02-06 00:50:39,622 INFO [optim.py:369] (0/4) 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 00:50:44,418 INFO [train.py:901] (0/4) Epoch 5, batch 3200, loss[loss=0.2616, simple_loss=0.3261, pruned_loss=0.09851, over 8255.00 frames. ], tot_loss[loss=0.2942, simple_loss=0.3541, pruned_loss=0.1171, over 1617383.52 frames. ], batch size: 22, lr: 1.51e-02, grad_scale: 8.0 +2023-02-06 00:50:54,830 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35548.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:51:09,455 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35569.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:51:20,160 INFO [train.py:901] (0/4) Epoch 5, batch 3250, loss[loss=0.3049, simple_loss=0.3716, pruned_loss=0.1191, over 8451.00 frames. ], tot_loss[loss=0.2957, simple_loss=0.3559, pruned_loss=0.1177, over 1621355.46 frames. ], batch size: 27, lr: 1.51e-02, grad_scale: 8.0 +2023-02-06 00:51:26,857 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 +2023-02-06 00:51:50,492 INFO [optim.py:369] (0/4) 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 00:51:55,280 INFO [train.py:901] (0/4) Epoch 5, batch 3300, loss[loss=0.2462, simple_loss=0.3096, pruned_loss=0.09136, over 7698.00 frames. ], tot_loss[loss=0.2922, simple_loss=0.3529, pruned_loss=0.1158, over 1617429.18 frames. ], batch size: 18, lr: 1.51e-02, grad_scale: 8.0 +2023-02-06 00:52:30,152 INFO [train.py:901] (0/4) Epoch 5, batch 3350, loss[loss=0.3105, simple_loss=0.3556, pruned_loss=0.1327, over 7812.00 frames. ], tot_loss[loss=0.2929, simple_loss=0.3533, pruned_loss=0.1162, over 1621556.89 frames. ], batch size: 20, lr: 1.51e-02, grad_scale: 16.0 +2023-02-06 00:52:47,810 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=35707.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 00:53:01,460 INFO [optim.py:369] (0/4) 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 00:53:06,222 INFO [train.py:901] (0/4) Epoch 5, batch 3400, loss[loss=0.2238, simple_loss=0.2903, pruned_loss=0.07865, over 8244.00 frames. ], tot_loss[loss=0.2926, simple_loss=0.353, pruned_loss=0.1161, over 1617918.54 frames. ], batch size: 22, lr: 1.51e-02, grad_scale: 16.0 +2023-02-06 00:53:08,383 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=35736.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:53:20,098 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6538, 2.1730, 4.3325, 1.1794, 2.9043, 2.2805, 1.5192, 2.3613], + device='cuda:0'), covar=tensor([0.1517, 0.2066, 0.0669, 0.3303, 0.1395, 0.2233, 0.1542, 0.2638], + device='cuda:0'), in_proj_covar=tensor([0.0465, 0.0443, 0.0523, 0.0531, 0.0573, 0.0512, 0.0446, 0.0590], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:0') +2023-02-06 00:53:39,717 INFO [train.py:901] (0/4) Epoch 5, batch 3450, loss[loss=0.2525, simple_loss=0.3159, pruned_loss=0.09453, over 8090.00 frames. ], tot_loss[loss=0.293, simple_loss=0.3533, pruned_loss=0.1164, over 1614052.11 frames. ], batch size: 21, lr: 1.51e-02, grad_scale: 16.0 +2023-02-06 00:53:43,879 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35789.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:53:45,410 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 +2023-02-06 00:54:01,268 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1864, 1.8306, 3.0702, 1.5940, 2.4386, 3.4502, 3.2495, 2.9742], + device='cuda:0'), covar=tensor([0.0829, 0.1208, 0.0490, 0.1669, 0.0742, 0.0235, 0.0398, 0.0578], + device='cuda:0'), in_proj_covar=tensor([0.0227, 0.0262, 0.0218, 0.0259, 0.0216, 0.0195, 0.0224, 0.0268], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 00:54:07,864 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35822.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 00:54:09,943 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35825.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:54:10,372 INFO [optim.py:369] (0/4) 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 00:54:12,575 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35829.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:54:15,092 INFO [train.py:901] (0/4) Epoch 5, batch 3500, loss[loss=0.2876, simple_loss=0.3592, pruned_loss=0.1081, over 8095.00 frames. ], tot_loss[loss=0.2945, simple_loss=0.3547, pruned_loss=0.1172, over 1616912.45 frames. ], batch size: 23, lr: 1.51e-02, grad_scale: 16.0 +2023-02-06 00:54:27,016 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35850.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:54:27,674 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35851.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:54:40,698 WARNING [train.py:1067] (0/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-06 00:54:40,871 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.6805, 2.9911, 2.4954, 3.9891, 2.0588, 2.0929, 2.3158, 3.2330], + device='cuda:0'), covar=tensor([0.1023, 0.1270, 0.1379, 0.0320, 0.1782, 0.2023, 0.2081, 0.1158], + device='cuda:0'), in_proj_covar=tensor([0.0270, 0.0267, 0.0287, 0.0223, 0.0252, 0.0277, 0.0290, 0.0264], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0004, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 00:54:48,819 INFO [train.py:901] (0/4) Epoch 5, batch 3550, loss[loss=0.2502, simple_loss=0.3061, pruned_loss=0.09713, over 7528.00 frames. ], tot_loss[loss=0.294, simple_loss=0.3541, pruned_loss=0.117, over 1614014.35 frames. ], batch size: 18, lr: 1.51e-02, grad_scale: 16.0 +2023-02-06 00:54:54,912 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=35892.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:55:16,695 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 +2023-02-06 00:55:19,528 INFO [optim.py:369] (0/4) 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 00:55:23,985 INFO [train.py:901] (0/4) Epoch 5, batch 3600, loss[loss=0.3276, simple_loss=0.376, pruned_loss=0.1396, over 8098.00 frames. ], tot_loss[loss=0.2946, simple_loss=0.3541, pruned_loss=0.1175, over 1613236.35 frames. ], batch size: 23, lr: 1.51e-02, grad_scale: 16.0 +2023-02-06 00:55:57,695 INFO [train.py:901] (0/4) Epoch 5, batch 3650, loss[loss=0.2961, simple_loss=0.3553, pruned_loss=0.1185, over 8791.00 frames. ], tot_loss[loss=0.2946, simple_loss=0.3542, pruned_loss=0.1175, over 1611813.30 frames. ], batch size: 30, lr: 1.51e-02, grad_scale: 16.0 +2023-02-06 00:56:08,724 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4695, 2.1479, 3.0199, 2.5849, 2.6316, 1.9935, 1.5275, 1.6436], + device='cuda:0'), covar=tensor([0.1321, 0.1526, 0.0386, 0.0765, 0.0660, 0.0866, 0.0915, 0.1479], + device='cuda:0'), in_proj_covar=tensor([0.0730, 0.0668, 0.0566, 0.0651, 0.0759, 0.0626, 0.0601, 0.0621], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 00:56:09,201 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-36000.pt +2023-02-06 00:56:15,053 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36007.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:56:27,680 INFO [optim.py:369] (0/4) 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 00:56:28,834 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.27 vs. limit=5.0 +2023-02-06 00:56:32,338 INFO [train.py:901] (0/4) Epoch 5, batch 3700, loss[loss=0.2847, simple_loss=0.3498, pruned_loss=0.1098, over 8444.00 frames. ], tot_loss[loss=0.2933, simple_loss=0.3533, pruned_loss=0.1166, over 1609001.93 frames. ], batch size: 27, lr: 1.50e-02, grad_scale: 16.0 +2023-02-06 00:56:40,798 WARNING [train.py:1067] (0/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-06 00:57:04,797 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36078.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 00:57:07,920 INFO [train.py:901] (0/4) Epoch 5, batch 3750, loss[loss=0.2431, simple_loss=0.3151, pruned_loss=0.08553, over 7808.00 frames. ], tot_loss[loss=0.294, simple_loss=0.3539, pruned_loss=0.117, over 1611516.32 frames. ], batch size: 20, lr: 1.50e-02, grad_scale: 16.0 +2023-02-06 00:57:14,130 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5154, 1.9997, 3.9330, 1.1196, 2.4432, 1.6187, 1.6447, 2.2756], + device='cuda:0'), covar=tensor([0.2149, 0.2549, 0.0721, 0.4170, 0.1877, 0.3286, 0.2182, 0.2807], + device='cuda:0'), in_proj_covar=tensor([0.0465, 0.0449, 0.0521, 0.0538, 0.0574, 0.0521, 0.0453, 0.0586], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:0') +2023-02-06 00:57:21,376 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36103.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 00:57:24,147 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36107.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:57:37,183 INFO [optim.py:369] (0/4) 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 00:57:40,766 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36132.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:57:41,236 INFO [train.py:901] (0/4) Epoch 5, batch 3800, loss[loss=0.2868, simple_loss=0.3464, pruned_loss=0.1136, over 8610.00 frames. ], tot_loss[loss=0.2932, simple_loss=0.3535, pruned_loss=0.1165, over 1610505.64 frames. ], batch size: 31, lr: 1.50e-02, grad_scale: 8.0 +2023-02-06 00:57:41,310 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=36133.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:58:05,376 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2060, 1.3605, 2.2554, 0.9569, 2.2362, 2.4638, 2.4022, 2.0710], + device='cuda:0'), covar=tensor([0.1020, 0.1026, 0.0474, 0.1923, 0.0487, 0.0384, 0.0543, 0.0761], + device='cuda:0'), in_proj_covar=tensor([0.0235, 0.0267, 0.0224, 0.0262, 0.0222, 0.0199, 0.0231, 0.0271], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 00:58:09,320 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=36173.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:58:15,710 INFO [train.py:901] (0/4) Epoch 5, batch 3850, loss[loss=0.2654, simple_loss=0.3204, pruned_loss=0.1052, over 7558.00 frames. ], tot_loss[loss=0.2946, simple_loss=0.3539, pruned_loss=0.1176, over 1606480.57 frames. ], batch size: 18, lr: 1.50e-02, grad_scale: 8.0 +2023-02-06 00:58:27,209 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) 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 00:58:49,706 INFO [train.py:901] (0/4) Epoch 5, batch 3900, loss[loss=0.252, simple_loss=0.3297, pruned_loss=0.08714, over 8299.00 frames. ], tot_loss[loss=0.2928, simple_loss=0.3532, pruned_loss=0.1162, over 1610584.43 frames. ], batch size: 23, lr: 1.50e-02, grad_scale: 8.0 +2023-02-06 00:58:59,592 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36248.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:59:02,912 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.8098, 1.3912, 5.8426, 2.0995, 5.1742, 4.9633, 5.3428, 5.3194], + device='cuda:0'), covar=tensor([0.0320, 0.3786, 0.0170, 0.2165, 0.0704, 0.0436, 0.0304, 0.0345], + device='cuda:0'), in_proj_covar=tensor([0.0336, 0.0498, 0.0415, 0.0420, 0.0491, 0.0407, 0.0398, 0.0451], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 00:59:09,628 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36263.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:59:24,818 INFO [train.py:901] (0/4) Epoch 5, batch 3950, loss[loss=0.2956, simple_loss=0.3581, pruned_loss=0.1165, over 8558.00 frames. ], tot_loss[loss=0.2894, simple_loss=0.3499, pruned_loss=0.1145, over 1606296.37 frames. ], batch size: 31, lr: 1.50e-02, grad_scale: 8.0 +2023-02-06 00:59:28,389 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36288.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:59:28,413 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36288.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 00:59:54,576 INFO [optim.py:369] (0/4) 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 00:59:58,379 INFO [train.py:901] (0/4) Epoch 5, batch 4000, loss[loss=0.2699, simple_loss=0.3232, pruned_loss=0.1083, over 7807.00 frames. ], tot_loss[loss=0.2913, simple_loss=0.3512, pruned_loss=0.1157, over 1611216.18 frames. ], batch size: 20, lr: 1.50e-02, grad_scale: 8.0 +2023-02-06 01:00:22,602 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.6906, 5.7973, 4.8943, 2.4581, 4.9799, 5.3557, 5.3407, 4.8632], + device='cuda:0'), covar=tensor([0.0735, 0.0422, 0.0873, 0.4118, 0.0608, 0.0726, 0.1097, 0.0659], + device='cuda:0'), in_proj_covar=tensor([0.0387, 0.0280, 0.0314, 0.0399, 0.0312, 0.0270, 0.0298, 0.0241], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], + device='cuda:0') +2023-02-06 01:00:31,743 INFO [train.py:901] (0/4) Epoch 5, batch 4050, loss[loss=0.2543, simple_loss=0.3301, pruned_loss=0.08919, over 8288.00 frames. ], tot_loss[loss=0.2922, simple_loss=0.352, pruned_loss=0.1162, over 1616164.57 frames. ], batch size: 23, lr: 1.50e-02, grad_scale: 8.0 +2023-02-06 01:01:03,217 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 5, batch 4100, loss[loss=0.2468, simple_loss=0.3061, pruned_loss=0.09373, over 7783.00 frames. ], tot_loss[loss=0.2901, simple_loss=0.3505, pruned_loss=0.1149, over 1616473.55 frames. ], batch size: 19, lr: 1.50e-02, grad_scale: 8.0 +2023-02-06 01:01:41,908 INFO [train.py:901] (0/4) Epoch 5, batch 4150, loss[loss=0.3142, simple_loss=0.3783, pruned_loss=0.125, over 8503.00 frames. ], tot_loss[loss=0.2902, simple_loss=0.3509, pruned_loss=0.1147, over 1620716.74 frames. ], batch size: 28, lr: 1.50e-02, grad_scale: 8.0 +2023-02-06 01:01:56,697 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36504.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:02:13,616 INFO [optim.py:369] (0/4) 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 01:02:15,198 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36529.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:02:17,751 INFO [train.py:901] (0/4) Epoch 5, batch 4200, loss[loss=0.2929, simple_loss=0.365, pruned_loss=0.1104, over 8638.00 frames. ], tot_loss[loss=0.289, simple_loss=0.3499, pruned_loss=0.114, over 1617021.24 frames. ], batch size: 34, lr: 1.49e-02, grad_scale: 8.0 +2023-02-06 01:02:24,845 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=36543.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:02:25,690 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36544.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:02:43,322 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-06 01:02:43,535 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36569.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:02:52,975 INFO [train.py:901] (0/4) Epoch 5, batch 4250, loss[loss=0.2801, simple_loss=0.3512, pruned_loss=0.1045, over 7969.00 frames. ], tot_loss[loss=0.2898, simple_loss=0.3506, pruned_loss=0.1145, over 1616698.65 frames. ], batch size: 21, lr: 1.49e-02, grad_scale: 8.0 +2023-02-06 01:03:05,722 WARNING [train.py:1067] (0/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-06 01:03:20,914 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36623.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:03:24,291 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 01:03:27,126 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36630.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:03:28,984 INFO [train.py:901] (0/4) Epoch 5, batch 4300, loss[loss=0.3352, simple_loss=0.3933, pruned_loss=0.1386, over 8616.00 frames. ], tot_loss[loss=0.2895, simple_loss=0.3502, pruned_loss=0.1144, over 1610724.12 frames. ], batch size: 39, lr: 1.49e-02, grad_scale: 8.0 +2023-02-06 01:03:47,191 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36658.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:04:05,007 INFO [train.py:901] (0/4) Epoch 5, batch 4350, loss[loss=0.2603, simple_loss=0.3323, pruned_loss=0.09418, over 7650.00 frames. ], tot_loss[loss=0.2886, simple_loss=0.3498, pruned_loss=0.1137, over 1611752.67 frames. ], batch size: 19, lr: 1.49e-02, grad_scale: 8.0 +2023-02-06 01:04:28,892 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36718.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 01:04:34,838 INFO [optim.py:369] (0/4) 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 01:04:36,272 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-06 01:04:38,962 INFO [train.py:901] (0/4) Epoch 5, batch 4400, loss[loss=0.2821, simple_loss=0.3512, pruned_loss=0.1065, over 8446.00 frames. ], tot_loss[loss=0.2901, simple_loss=0.351, pruned_loss=0.1146, over 1610354.11 frames. ], batch size: 27, lr: 1.49e-02, grad_scale: 8.0 +2023-02-06 01:05:15,305 INFO [train.py:901] (0/4) Epoch 5, batch 4450, loss[loss=0.2948, simple_loss=0.3619, pruned_loss=0.1139, over 8316.00 frames. ], tot_loss[loss=0.2907, simple_loss=0.3514, pruned_loss=0.115, over 1612627.32 frames. ], batch size: 25, lr: 1.49e-02, grad_scale: 8.0 +2023-02-06 01:05:18,095 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-06 01:05:43,210 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36823.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 01:05:45,712 INFO [optim.py:369] (0/4) 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 01:05:49,707 INFO [train.py:901] (0/4) Epoch 5, batch 4500, loss[loss=0.3179, simple_loss=0.3669, pruned_loss=0.1344, over 8682.00 frames. ], tot_loss[loss=0.2884, simple_loss=0.3489, pruned_loss=0.1139, over 1609490.54 frames. ], batch size: 39, lr: 1.49e-02, grad_scale: 8.0 +2023-02-06 01:06:15,008 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-06 01:06:25,784 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 +2023-02-06 01:06:26,116 INFO [train.py:901] (0/4) Epoch 5, batch 4550, loss[loss=0.2288, simple_loss=0.2922, pruned_loss=0.08267, over 7539.00 frames. ], tot_loss[loss=0.2883, simple_loss=0.349, pruned_loss=0.1138, over 1606410.36 frames. ], batch size: 18, lr: 1.49e-02, grad_scale: 8.0 +2023-02-06 01:06:47,949 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 01:06:58,101 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 5, batch 4600, loss[loss=0.3309, simple_loss=0.3745, pruned_loss=0.1437, over 6719.00 frames. ], tot_loss[loss=0.2887, simple_loss=0.3493, pruned_loss=0.1141, over 1609144.86 frames. ], batch size: 71, lr: 1.49e-02, grad_scale: 8.0 +2023-02-06 01:07:04,720 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36939.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:07:23,544 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=36967.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:07:25,626 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=36974.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:07:30,326 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36976.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:07:35,634 INFO [train.py:901] (0/4) Epoch 5, batch 4650, loss[loss=0.2597, simple_loss=0.3285, pruned_loss=0.09549, over 7537.00 frames. ], tot_loss[loss=0.2879, simple_loss=0.3482, pruned_loss=0.1138, over 1609596.61 frames. ], batch size: 18, lr: 1.49e-02, grad_scale: 8.0 +2023-02-06 01:07:53,422 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3058, 1.5353, 1.2765, 1.9301, 0.8285, 1.2181, 1.2620, 1.4419], + device='cuda:0'), covar=tensor([0.1233, 0.1209, 0.1642, 0.0623, 0.1559, 0.2002, 0.1155, 0.1030], + device='cuda:0'), in_proj_covar=tensor([0.0275, 0.0269, 0.0294, 0.0227, 0.0259, 0.0291, 0.0289, 0.0266], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0004, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 01:08:02,858 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37022.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:08:06,068 INFO [optim.py:369] (0/4) 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 01:08:10,729 INFO [train.py:901] (0/4) Epoch 5, batch 4700, loss[loss=0.2312, simple_loss=0.2985, pruned_loss=0.08198, over 7812.00 frames. ], tot_loss[loss=0.2871, simple_loss=0.3472, pruned_loss=0.1135, over 1603049.21 frames. ], batch size: 19, lr: 1.48e-02, grad_scale: 8.0 +2023-02-06 01:08:16,882 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6412, 1.1515, 1.3785, 1.0578, 1.0221, 1.2083, 1.3157, 1.2841], + device='cuda:0'), covar=tensor([0.0613, 0.1427, 0.1864, 0.1596, 0.0637, 0.1679, 0.0797, 0.0590], + device='cuda:0'), in_proj_covar=tensor([0.0138, 0.0181, 0.0222, 0.0183, 0.0132, 0.0191, 0.0147, 0.0156], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0007, 0.0008, 0.0007, 0.0005, 0.0007, 0.0006, 0.0006], + device='cuda:0') +2023-02-06 01:08:29,882 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=37062.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 01:08:43,582 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37082.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:08:44,112 INFO [train.py:901] (0/4) Epoch 5, batch 4750, loss[loss=0.2275, simple_loss=0.301, pruned_loss=0.07695, over 8234.00 frames. ], tot_loss[loss=0.2883, simple_loss=0.3484, pruned_loss=0.1141, over 1606065.66 frames. ], batch size: 22, lr: 1.48e-02, grad_scale: 8.0 +2023-02-06 01:08:45,690 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37085.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:08:49,119 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37089.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:09:08,545 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.2080, 3.1101, 2.9177, 1.5452, 2.8767, 2.8234, 2.9328, 2.5916], + device='cuda:0'), covar=tensor([0.1208, 0.0766, 0.1087, 0.4497, 0.1036, 0.0975, 0.1349, 0.0960], + device='cuda:0'), in_proj_covar=tensor([0.0399, 0.0290, 0.0324, 0.0406, 0.0323, 0.0270, 0.0305, 0.0247], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 01:09:15,976 INFO [optim.py:369] (0/4) 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 01:09:17,396 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-06 01:09:20,173 INFO [train.py:901] (0/4) Epoch 5, batch 4800, loss[loss=0.2837, simple_loss=0.3549, pruned_loss=0.1062, over 8361.00 frames. ], tot_loss[loss=0.2877, simple_loss=0.3482, pruned_loss=0.1135, over 1610309.96 frames. ], batch size: 24, lr: 1.48e-02, grad_scale: 8.0 +2023-02-06 01:09:20,179 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-06 01:09:44,350 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=37167.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 01:09:51,274 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37177.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 01:09:51,920 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2380, 1.3277, 2.1241, 1.0883, 2.0886, 2.2879, 2.2824, 1.9662], + device='cuda:0'), covar=tensor([0.1049, 0.1143, 0.0612, 0.1871, 0.0598, 0.0412, 0.0563, 0.0800], + device='cuda:0'), in_proj_covar=tensor([0.0235, 0.0270, 0.0225, 0.0262, 0.0225, 0.0198, 0.0231, 0.0273], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 01:09:55,225 INFO [train.py:901] (0/4) Epoch 5, batch 4850, loss[loss=0.2646, simple_loss=0.3226, pruned_loss=0.1033, over 7970.00 frames. ], tot_loss[loss=0.2873, simple_loss=0.3481, pruned_loss=0.1132, over 1609988.45 frames. ], batch size: 21, lr: 1.48e-02, grad_scale: 8.0 +2023-02-06 01:10:10,661 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-06 01:10:12,992 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=2.01 vs. limit=2.0 +2023-02-06 01:10:27,459 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 5, batch 4900, loss[loss=0.2668, simple_loss=0.3447, pruned_loss=0.09444, over 8448.00 frames. ], tot_loss[loss=0.2872, simple_loss=0.3483, pruned_loss=0.113, over 1612761.51 frames. ], batch size: 27, lr: 1.48e-02, grad_scale: 8.0 +2023-02-06 01:10:59,372 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=37273.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:11:05,592 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37282.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 01:11:06,080 INFO [train.py:901] (0/4) Epoch 5, batch 4950, loss[loss=0.2733, simple_loss=0.3419, pruned_loss=0.1024, over 8470.00 frames. ], tot_loss[loss=0.2858, simple_loss=0.3475, pruned_loss=0.112, over 1611177.46 frames. ], batch size: 28, lr: 1.48e-02, grad_scale: 8.0 +2023-02-06 01:11:08,182 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37286.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:11:17,614 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.1775, 1.0983, 1.1350, 1.0937, 0.7728, 1.1597, 0.0280, 0.8678], + device='cuda:0'), covar=tensor([0.3440, 0.2357, 0.1179, 0.2065, 0.6705, 0.1062, 0.5514, 0.2055], + device='cuda:0'), in_proj_covar=tensor([0.0129, 0.0124, 0.0078, 0.0173, 0.0209, 0.0082, 0.0150, 0.0126], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 01:11:31,041 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) Epoch 5, batch 5000, loss[loss=0.2695, simple_loss=0.3245, pruned_loss=0.1073, over 7201.00 frames. ], tot_loss[loss=0.286, simple_loss=0.3474, pruned_loss=0.1123, over 1608064.85 frames. ], batch size: 16, lr: 1.48e-02, grad_scale: 8.0 +2023-02-06 01:11:43,908 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37338.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:11:45,908 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37341.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:11:49,228 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37345.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:11:55,335 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.7729, 2.9550, 3.0444, 1.9314, 1.3914, 3.1781, 0.4620, 1.6937], + device='cuda:0'), covar=tensor([0.2442, 0.1252, 0.0960, 0.3851, 0.6393, 0.0539, 0.7023, 0.3071], + device='cuda:0'), in_proj_covar=tensor([0.0130, 0.0125, 0.0078, 0.0174, 0.0209, 0.0082, 0.0150, 0.0126], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 01:12:01,321 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37363.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:12:02,610 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2286, 1.4495, 1.2316, 1.8703, 0.9248, 1.0694, 1.2225, 1.4208], + device='cuda:0'), covar=tensor([0.1201, 0.1209, 0.1556, 0.0746, 0.1739, 0.2390, 0.1397, 0.1057], + device='cuda:0'), in_proj_covar=tensor([0.0273, 0.0267, 0.0291, 0.0229, 0.0259, 0.0289, 0.0294, 0.0265], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0004, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 01:12:03,189 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=37366.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:12:03,304 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37366.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:12:05,923 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37370.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:12:13,775 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37382.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:12:15,009 INFO [train.py:901] (0/4) Epoch 5, batch 5050, loss[loss=0.2849, simple_loss=0.3565, pruned_loss=0.1066, over 8323.00 frames. ], tot_loss[loss=0.2871, simple_loss=0.349, pruned_loss=0.1126, over 1615362.93 frames. ], batch size: 25, lr: 1.48e-02, grad_scale: 8.0 +2023-02-06 01:12:18,395 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37388.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:12:44,032 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7859, 1.8657, 2.1764, 1.6042, 0.9401, 2.1842, 0.3874, 1.0659], + device='cuda:0'), covar=tensor([0.2991, 0.2543, 0.0860, 0.2560, 0.6281, 0.0687, 0.5661, 0.2547], + device='cuda:0'), in_proj_covar=tensor([0.0133, 0.0127, 0.0079, 0.0175, 0.0211, 0.0083, 0.0151, 0.0128], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 01:12:44,482 INFO [optim.py:369] (0/4) 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 01:12:48,495 INFO [train.py:901] (0/4) Epoch 5, batch 5100, loss[loss=0.2302, simple_loss=0.2937, pruned_loss=0.08336, over 7438.00 frames. ], tot_loss[loss=0.2866, simple_loss=0.3487, pruned_loss=0.1122, over 1612339.75 frames. ], batch size: 17, lr: 1.48e-02, grad_scale: 8.0 +2023-02-06 01:12:48,506 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-06 01:12:48,711 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37433.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 01:12:49,914 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37435.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:12:53,990 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.1100, 1.0823, 1.1098, 1.1127, 0.7794, 1.2253, 0.0347, 0.8986], + device='cuda:0'), covar=tensor([0.3177, 0.2677, 0.1176, 0.1999, 0.6322, 0.0859, 0.5069, 0.2162], + device='cuda:0'), in_proj_covar=tensor([0.0132, 0.0127, 0.0079, 0.0174, 0.0210, 0.0083, 0.0151, 0.0127], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 01:13:06,528 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37458.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 01:13:15,615 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.6922, 1.2012, 3.8233, 1.3591, 3.2757, 3.2103, 3.4242, 3.3869], + device='cuda:0'), covar=tensor([0.0480, 0.3517, 0.0425, 0.2638, 0.1109, 0.0646, 0.0530, 0.0599], + device='cuda:0'), in_proj_covar=tensor([0.0336, 0.0497, 0.0417, 0.0421, 0.0487, 0.0411, 0.0404, 0.0453], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 01:13:16,275 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8620, 1.5181, 3.0598, 1.2526, 2.2276, 3.2931, 3.2014, 2.7866], + device='cuda:0'), covar=tensor([0.0958, 0.1401, 0.0406, 0.1997, 0.0673, 0.0275, 0.0518, 0.0716], + device='cuda:0'), in_proj_covar=tensor([0.0239, 0.0277, 0.0228, 0.0266, 0.0227, 0.0200, 0.0233, 0.0281], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 01:13:22,242 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 5, batch 5150, loss[loss=0.2796, simple_loss=0.3442, pruned_loss=0.1075, over 8027.00 frames. ], tot_loss[loss=0.2884, simple_loss=0.3501, pruned_loss=0.1133, over 1609955.77 frames. ], batch size: 22, lr: 1.48e-02, grad_scale: 8.0 +2023-02-06 01:13:28,193 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-06 01:13:33,446 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.7256, 3.1723, 2.4446, 4.0805, 2.0552, 2.1512, 2.3036, 3.1756], + device='cuda:0'), covar=tensor([0.0859, 0.1030, 0.1305, 0.0256, 0.1451, 0.1917, 0.1623, 0.1037], + device='cuda:0'), in_proj_covar=tensor([0.0274, 0.0265, 0.0292, 0.0227, 0.0255, 0.0291, 0.0291, 0.0264], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0004, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 01:13:53,498 INFO [optim.py:369] (0/4) 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 01:13:57,560 INFO [train.py:901] (0/4) Epoch 5, batch 5200, loss[loss=0.2794, simple_loss=0.3418, pruned_loss=0.1085, over 8247.00 frames. ], tot_loss[loss=0.2872, simple_loss=0.3489, pruned_loss=0.1128, over 1610070.55 frames. ], batch size: 24, lr: 1.47e-02, grad_scale: 8.0 +2023-02-06 01:14:01,122 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37538.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 01:14:19,384 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37563.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 01:14:33,081 INFO [train.py:901] (0/4) Epoch 5, batch 5250, loss[loss=0.2854, simple_loss=0.3319, pruned_loss=0.1194, over 7707.00 frames. ], tot_loss[loss=0.2873, simple_loss=0.3489, pruned_loss=0.1128, over 1610278.22 frames. ], batch size: 18, lr: 1.47e-02, grad_scale: 8.0 +2023-02-06 01:14:45,217 WARNING [train.py:1067] (0/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-06 01:14:57,683 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9531, 2.4649, 1.8912, 3.0641, 1.5804, 1.8072, 1.8789, 2.5215], + device='cuda:0'), covar=tensor([0.1102, 0.1121, 0.1388, 0.0394, 0.1463, 0.1876, 0.1687, 0.1083], + device='cuda:0'), in_proj_covar=tensor([0.0269, 0.0260, 0.0284, 0.0220, 0.0250, 0.0281, 0.0287, 0.0259], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0004, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 01:15:03,493 INFO [optim.py:369] (0/4) 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 01:15:05,547 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=37630.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:15:07,536 INFO [train.py:901] (0/4) Epoch 5, batch 5300, loss[loss=0.337, simple_loss=0.387, pruned_loss=0.1434, over 8598.00 frames. ], tot_loss[loss=0.2881, simple_loss=0.3493, pruned_loss=0.1134, over 1607796.66 frames. ], batch size: 34, lr: 1.47e-02, grad_scale: 8.0 +2023-02-06 01:15:15,186 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37644.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:15:32,580 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37669.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:15:42,448 INFO [train.py:901] (0/4) Epoch 5, batch 5350, loss[loss=0.2381, simple_loss=0.3079, pruned_loss=0.08416, over 7715.00 frames. ], tot_loss[loss=0.288, simple_loss=0.3493, pruned_loss=0.1134, over 1610320.70 frames. ], batch size: 18, lr: 1.47e-02, grad_scale: 8.0 +2023-02-06 01:15:47,815 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37691.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:16:05,060 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) 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] (0/4) Epoch 5, batch 5400, loss[loss=0.2908, simple_loss=0.353, pruned_loss=0.1143, over 8037.00 frames. ], tot_loss[loss=0.2882, simple_loss=0.349, pruned_loss=0.1137, over 1609855.06 frames. ], batch size: 22, lr: 1.47e-02, grad_scale: 8.0 +2023-02-06 01:16:19,889 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37737.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:16:25,206 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37745.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:16:36,759 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37762.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:16:50,837 INFO [train.py:901] (0/4) Epoch 5, batch 5450, loss[loss=0.2717, simple_loss=0.3464, pruned_loss=0.09857, over 8255.00 frames. ], tot_loss[loss=0.2894, simple_loss=0.3501, pruned_loss=0.1143, over 1608398.29 frames. ], batch size: 24, lr: 1.47e-02, grad_scale: 8.0 +2023-02-06 01:17:22,465 INFO [optim.py:369] (0/4) 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 01:17:23,290 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9336, 2.3741, 4.8327, 1.3257, 3.1261, 2.5816, 1.9373, 2.5685], + device='cuda:0'), covar=tensor([0.1227, 0.1620, 0.0528, 0.2892, 0.1334, 0.1887, 0.1239, 0.2331], + device='cuda:0'), in_proj_covar=tensor([0.0462, 0.0439, 0.0521, 0.0530, 0.0577, 0.0507, 0.0442, 0.0586], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003], + device='cuda:0') +2023-02-06 01:17:26,495 INFO [train.py:901] (0/4) Epoch 5, batch 5500, loss[loss=0.2948, simple_loss=0.3545, pruned_loss=0.1176, over 7972.00 frames. ], tot_loss[loss=0.2895, simple_loss=0.3501, pruned_loss=0.1145, over 1609222.38 frames. ], batch size: 21, lr: 1.47e-02, grad_scale: 8.0 +2023-02-06 01:17:30,479 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-06 01:17:31,993 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37841.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:18:00,735 INFO [train.py:901] (0/4) Epoch 5, batch 5550, loss[loss=0.2516, simple_loss=0.3144, pruned_loss=0.09438, over 7815.00 frames. ], tot_loss[loss=0.2883, simple_loss=0.3492, pruned_loss=0.1137, over 1605586.76 frames. ], batch size: 20, lr: 1.47e-02, grad_scale: 8.0 +2023-02-06 01:18:13,690 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4322, 1.9340, 2.1371, 0.9370, 2.1242, 1.3455, 0.5072, 1.7164], + device='cuda:0'), covar=tensor([0.0244, 0.0104, 0.0063, 0.0190, 0.0127, 0.0345, 0.0340, 0.0092], + device='cuda:0'), in_proj_covar=tensor([0.0313, 0.0225, 0.0183, 0.0267, 0.0215, 0.0362, 0.0284, 0.0258], + device='cuda:0'), out_proj_covar=tensor([1.1242e-04, 7.8658e-05, 6.2861e-05, 9.3047e-05, 7.7154e-05, 1.3821e-04, + 1.0211e-04, 9.0575e-05], device='cuda:0') +2023-02-06 01:18:19,135 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.1349, 2.0122, 1.9559, 1.8361, 1.7189, 1.7770, 2.7393, 2.4545], + device='cuda:0'), covar=tensor([0.0784, 0.1891, 0.2809, 0.1872, 0.0768, 0.2324, 0.0820, 0.0727], + device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0180, 0.0219, 0.0181, 0.0131, 0.0191, 0.0146, 0.0154], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0007, 0.0008, 0.0007, 0.0005, 0.0007, 0.0006, 0.0006], + device='cuda:0') +2023-02-06 01:18:32,126 INFO [optim.py:369] (0/4) 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 01:18:36,192 INFO [train.py:901] (0/4) Epoch 5, batch 5600, loss[loss=0.321, simple_loss=0.3768, pruned_loss=0.1326, over 8615.00 frames. ], tot_loss[loss=0.2881, simple_loss=0.3491, pruned_loss=0.1136, over 1607697.60 frames. ], batch size: 31, lr: 1.47e-02, grad_scale: 8.0 +2023-02-06 01:18:53,364 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 +2023-02-06 01:19:10,750 INFO [train.py:901] (0/4) Epoch 5, batch 5650, loss[loss=0.3041, simple_loss=0.364, pruned_loss=0.1221, over 8340.00 frames. ], tot_loss[loss=0.2884, simple_loss=0.3495, pruned_loss=0.1137, over 1607209.59 frames. ], batch size: 25, lr: 1.47e-02, grad_scale: 4.0 +2023-02-06 01:19:20,215 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37997.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:19:22,188 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-38000.pt +2023-02-06 01:19:23,901 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=38001.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:19:34,405 WARNING [train.py:1067] (0/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-06 01:19:41,416 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=38026.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:19:42,548 INFO [optim.py:369] (0/4) 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 01:19:45,906 INFO [train.py:901] (0/4) Epoch 5, batch 5700, loss[loss=0.246, simple_loss=0.318, pruned_loss=0.08707, over 8517.00 frames. ], tot_loss[loss=0.2875, simple_loss=0.3489, pruned_loss=0.1131, over 1611830.00 frames. ], batch size: 26, lr: 1.46e-02, grad_scale: 4.0 +2023-02-06 01:20:20,747 INFO [train.py:901] (0/4) Epoch 5, batch 5750, loss[loss=0.2016, simple_loss=0.2717, pruned_loss=0.06571, over 7426.00 frames. ], tot_loss[loss=0.2899, simple_loss=0.3509, pruned_loss=0.1144, over 1610963.26 frames. ], batch size: 17, lr: 1.46e-02, grad_scale: 4.0 +2023-02-06 01:20:27,208 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.17 vs. limit=5.0 +2023-02-06 01:20:30,247 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=38097.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:20:37,193 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-06 01:20:46,788 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=38122.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:20:50,577 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 5, batch 5800, loss[loss=0.2537, simple_loss=0.323, pruned_loss=0.09223, over 7804.00 frames. ], tot_loss[loss=0.2893, simple_loss=0.3504, pruned_loss=0.1141, over 1610615.09 frames. ], batch size: 20, lr: 1.46e-02, grad_scale: 4.0 +2023-02-06 01:21:03,327 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6319, 1.5626, 3.0408, 1.1990, 2.1526, 3.2439, 3.3117, 2.8220], + device='cuda:0'), covar=tensor([0.1075, 0.1289, 0.0373, 0.1883, 0.0707, 0.0298, 0.0401, 0.0679], + device='cuda:0'), in_proj_covar=tensor([0.0237, 0.0269, 0.0225, 0.0263, 0.0225, 0.0201, 0.0232, 0.0277], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 01:21:29,412 INFO [train.py:901] (0/4) Epoch 5, batch 5850, loss[loss=0.2997, simple_loss=0.3562, pruned_loss=0.1216, over 7820.00 frames. ], tot_loss[loss=0.2895, simple_loss=0.3502, pruned_loss=0.1144, over 1607733.63 frames. ], batch size: 20, lr: 1.46e-02, grad_scale: 4.0 +2023-02-06 01:22:01,285 INFO [optim.py:369] (0/4) 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 01:22:04,817 INFO [train.py:901] (0/4) Epoch 5, batch 5900, loss[loss=0.2752, simple_loss=0.3479, pruned_loss=0.1013, over 8239.00 frames. ], tot_loss[loss=0.2888, simple_loss=0.3495, pruned_loss=0.114, over 1607358.88 frames. ], batch size: 22, lr: 1.46e-02, grad_scale: 4.0 +2023-02-06 01:22:25,340 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4191, 1.4899, 1.5173, 1.2494, 1.2254, 1.4883, 1.6724, 1.4987], + device='cuda:0'), covar=tensor([0.0552, 0.1283, 0.1941, 0.1489, 0.0649, 0.1616, 0.0821, 0.0610], + device='cuda:0'), in_proj_covar=tensor([0.0134, 0.0177, 0.0217, 0.0181, 0.0129, 0.0187, 0.0142, 0.0152], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0007, 0.0008, 0.0007, 0.0005, 0.0007, 0.0006, 0.0006], + device='cuda:0') +2023-02-06 01:22:40,895 INFO [train.py:901] (0/4) Epoch 5, batch 5950, loss[loss=0.2814, simple_loss=0.3435, pruned_loss=0.1096, over 8143.00 frames. ], tot_loss[loss=0.2896, simple_loss=0.3501, pruned_loss=0.1146, over 1609008.09 frames. ], batch size: 22, lr: 1.46e-02, grad_scale: 4.0 +2023-02-06 01:22:55,068 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.1218, 3.0982, 2.8611, 1.4614, 2.7639, 2.7802, 2.9637, 2.6033], + device='cuda:0'), covar=tensor([0.1405, 0.0832, 0.1152, 0.4894, 0.1112, 0.1069, 0.1402, 0.1116], + device='cuda:0'), in_proj_covar=tensor([0.0403, 0.0291, 0.0320, 0.0409, 0.0320, 0.0272, 0.0305, 0.0252], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 01:23:12,021 INFO [optim.py:369] (0/4) 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 01:23:15,479 INFO [train.py:901] (0/4) Epoch 5, batch 6000, loss[loss=0.2557, simple_loss=0.3228, pruned_loss=0.09426, over 8491.00 frames. ], tot_loss[loss=0.289, simple_loss=0.3495, pruned_loss=0.1142, over 1609157.44 frames. ], batch size: 29, lr: 1.46e-02, grad_scale: 8.0 +2023-02-06 01:23:15,479 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 01:23:28,273 INFO [train.py:935] (0/4) Epoch 5, validation: loss=0.2196, simple_loss=0.3162, pruned_loss=0.06146, over 944034.00 frames. +2023-02-06 01:23:28,274 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6689MB +2023-02-06 01:23:33,797 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=38341.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:23:55,755 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5810, 1.9679, 1.6517, 2.4197, 1.0188, 1.3277, 1.5402, 1.7996], + device='cuda:0'), covar=tensor([0.1221, 0.1218, 0.1611, 0.0628, 0.1728, 0.2226, 0.1553, 0.1229], + device='cuda:0'), in_proj_covar=tensor([0.0271, 0.0256, 0.0278, 0.0218, 0.0248, 0.0281, 0.0291, 0.0263], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0004, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 01:23:58,478 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=38378.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:24:01,709 INFO [train.py:901] (0/4) Epoch 5, batch 6050, loss[loss=0.3499, simple_loss=0.4051, pruned_loss=0.1474, over 8461.00 frames. ], tot_loss[loss=0.2906, simple_loss=0.3505, pruned_loss=0.1154, over 1613333.87 frames. ], batch size: 25, lr: 1.46e-02, grad_scale: 8.0 +2023-02-06 01:24:12,375 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8028, 2.0862, 2.2153, 1.8206, 1.0848, 2.1843, 0.3499, 1.2866], + device='cuda:0'), covar=tensor([0.3023, 0.2036, 0.1095, 0.2398, 0.6939, 0.0845, 0.5734, 0.2658], + device='cuda:0'), in_proj_covar=tensor([0.0128, 0.0121, 0.0082, 0.0167, 0.0211, 0.0080, 0.0140, 0.0124], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 01:24:26,789 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.76 vs. limit=5.0 +2023-02-06 01:24:33,771 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 5, batch 6100, loss[loss=0.2493, simple_loss=0.3049, pruned_loss=0.09688, over 7662.00 frames. ], tot_loss[loss=0.2897, simple_loss=0.3497, pruned_loss=0.1149, over 1610014.82 frames. ], batch size: 19, lr: 1.46e-02, grad_scale: 8.0 +2023-02-06 01:24:53,384 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=38456.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:25:09,667 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-06 01:25:10,931 INFO [train.py:901] (0/4) Epoch 5, batch 6150, loss[loss=0.2735, simple_loss=0.3412, pruned_loss=0.1029, over 8246.00 frames. ], tot_loss[loss=0.2894, simple_loss=0.3498, pruned_loss=0.1145, over 1612078.57 frames. ], batch size: 24, lr: 1.46e-02, grad_scale: 8.0 +2023-02-06 01:25:42,283 INFO [optim.py:369] (0/4) 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 01:25:45,624 INFO [train.py:901] (0/4) Epoch 5, batch 6200, loss[loss=0.2893, simple_loss=0.3305, pruned_loss=0.1241, over 7439.00 frames. ], tot_loss[loss=0.2886, simple_loss=0.3492, pruned_loss=0.114, over 1611390.20 frames. ], batch size: 17, lr: 1.46e-02, grad_scale: 8.0 +2023-02-06 01:25:50,627 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 +2023-02-06 01:26:03,318 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-02-06 01:26:20,108 INFO [train.py:901] (0/4) Epoch 5, batch 6250, loss[loss=0.2627, simple_loss=0.3369, pruned_loss=0.09424, over 8286.00 frames. ], tot_loss[loss=0.289, simple_loss=0.3498, pruned_loss=0.1141, over 1616438.96 frames. ], batch size: 23, lr: 1.45e-02, grad_scale: 8.0 +2023-02-06 01:26:51,168 INFO [optim.py:369] (0/4) 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 01:26:54,590 INFO [train.py:901] (0/4) Epoch 5, batch 6300, loss[loss=0.2704, simple_loss=0.3433, pruned_loss=0.09881, over 8116.00 frames. ], tot_loss[loss=0.2888, simple_loss=0.3494, pruned_loss=0.1141, over 1616701.92 frames. ], batch size: 23, lr: 1.45e-02, grad_scale: 8.0 +2023-02-06 01:27:29,604 INFO [train.py:901] (0/4) Epoch 5, batch 6350, loss[loss=0.3467, simple_loss=0.387, pruned_loss=0.1532, over 8467.00 frames. ], tot_loss[loss=0.2878, simple_loss=0.3485, pruned_loss=0.1136, over 1611495.89 frames. ], batch size: 29, lr: 1.45e-02, grad_scale: 8.0 +2023-02-06 01:27:32,386 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.2373, 4.2706, 3.8279, 1.7257, 3.6631, 3.7564, 4.0257, 3.4014], + device='cuda:0'), covar=tensor([0.0806, 0.0573, 0.0893, 0.4898, 0.0840, 0.0885, 0.1028, 0.0729], + device='cuda:0'), in_proj_covar=tensor([0.0389, 0.0293, 0.0318, 0.0400, 0.0309, 0.0271, 0.0301, 0.0250], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 01:27:43,123 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3335, 1.8453, 2.9392, 2.4090, 2.5850, 1.8808, 1.4679, 1.3397], + device='cuda:0'), covar=tensor([0.1587, 0.2064, 0.0499, 0.0993, 0.0790, 0.1079, 0.1027, 0.1955], + device='cuda:0'), in_proj_covar=tensor([0.0755, 0.0684, 0.0580, 0.0663, 0.0781, 0.0640, 0.0611, 0.0635], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 01:27:49,908 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=38712.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:27:56,415 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=38722.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:28:00,326 INFO [optim.py:369] (0/4) 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 01:28:03,610 INFO [train.py:901] (0/4) Epoch 5, batch 6400, loss[loss=0.326, simple_loss=0.3705, pruned_loss=0.1407, over 8101.00 frames. ], tot_loss[loss=0.289, simple_loss=0.3499, pruned_loss=0.1141, over 1616275.33 frames. ], batch size: 23, lr: 1.45e-02, grad_scale: 8.0 +2023-02-06 01:28:06,523 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=38737.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:28:17,088 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.7513, 3.7119, 3.4210, 1.7011, 3.4125, 3.1918, 3.4899, 3.0867], + device='cuda:0'), covar=tensor([0.1163, 0.0741, 0.1073, 0.4946, 0.0829, 0.1285, 0.1394, 0.1186], + device='cuda:0'), in_proj_covar=tensor([0.0400, 0.0296, 0.0323, 0.0407, 0.0310, 0.0273, 0.0305, 0.0254], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 01:28:32,976 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.8879, 1.4962, 4.2905, 1.9228, 3.1409, 3.3854, 3.7982, 3.8065], + device='cuda:0'), covar=tensor([0.1112, 0.5551, 0.0871, 0.3312, 0.2247, 0.1275, 0.1050, 0.1132], + device='cuda:0'), in_proj_covar=tensor([0.0338, 0.0488, 0.0424, 0.0429, 0.0484, 0.0405, 0.0405, 0.0450], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 01:28:36,397 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1095, 2.2668, 2.8437, 0.9678, 2.8974, 1.8920, 1.3231, 1.6507], + device='cuda:0'), covar=tensor([0.0252, 0.0170, 0.0104, 0.0288, 0.0136, 0.0312, 0.0367, 0.0187], + device='cuda:0'), in_proj_covar=tensor([0.0318, 0.0227, 0.0192, 0.0274, 0.0220, 0.0365, 0.0286, 0.0264], + device='cuda:0'), out_proj_covar=tensor([1.1291e-04, 7.8632e-05, 6.5704e-05, 9.4632e-05, 7.7646e-05, 1.3749e-04, + 1.0192e-04, 9.2356e-05], device='cuda:0') +2023-02-06 01:28:38,890 INFO [train.py:901] (0/4) Epoch 5, batch 6450, loss[loss=0.354, simple_loss=0.3898, pruned_loss=0.1591, over 6522.00 frames. ], tot_loss[loss=0.2886, simple_loss=0.3496, pruned_loss=0.1138, over 1615813.87 frames. ], batch size: 71, lr: 1.45e-02, grad_scale: 8.0 +2023-02-06 01:28:57,430 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9164, 2.2573, 3.0168, 0.9735, 2.8648, 1.7876, 1.4327, 1.5850], + device='cuda:0'), covar=tensor([0.0274, 0.0132, 0.0066, 0.0267, 0.0125, 0.0317, 0.0354, 0.0184], + device='cuda:0'), in_proj_covar=tensor([0.0318, 0.0225, 0.0191, 0.0273, 0.0220, 0.0363, 0.0286, 0.0263], + device='cuda:0'), out_proj_covar=tensor([1.1293e-04, 7.7755e-05, 6.5417e-05, 9.4386e-05, 7.7793e-05, 1.3670e-04, + 1.0163e-04, 9.2006e-05], device='cuda:0') +2023-02-06 01:29:09,816 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 5, batch 6500, loss[loss=0.3105, simple_loss=0.3697, pruned_loss=0.1256, over 8338.00 frames. ], tot_loss[loss=0.2895, simple_loss=0.3508, pruned_loss=0.1141, over 1617265.24 frames. ], batch size: 26, lr: 1.45e-02, grad_scale: 8.0 +2023-02-06 01:29:16,020 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=38837.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:29:18,630 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=38841.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:29:30,093 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 +2023-02-06 01:29:48,032 INFO [train.py:901] (0/4) Epoch 5, batch 6550, loss[loss=0.3278, simple_loss=0.3856, pruned_loss=0.135, over 8670.00 frames. ], tot_loss[loss=0.2899, simple_loss=0.3513, pruned_loss=0.1143, over 1618501.39 frames. ], batch size: 34, lr: 1.45e-02, grad_scale: 8.0 +2023-02-06 01:30:04,623 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5109, 2.4317, 1.6742, 2.0409, 2.0461, 1.5163, 1.8059, 1.9534], + device='cuda:0'), covar=tensor([0.0889, 0.0283, 0.0754, 0.0428, 0.0497, 0.0998, 0.0660, 0.0550], + device='cuda:0'), in_proj_covar=tensor([0.0345, 0.0236, 0.0306, 0.0299, 0.0317, 0.0312, 0.0336, 0.0309], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 01:30:19,276 INFO [optim.py:369] (0/4) 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 01:30:22,026 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-06 01:30:22,684 INFO [train.py:901] (0/4) Epoch 5, batch 6600, loss[loss=0.3183, simple_loss=0.3785, pruned_loss=0.129, over 8444.00 frames. ], tot_loss[loss=0.2889, simple_loss=0.3503, pruned_loss=0.1137, over 1618295.89 frames. ], batch size: 27, lr: 1.45e-02, grad_scale: 8.0 +2023-02-06 01:30:39,809 WARNING [train.py:1067] (0/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-06 01:30:58,105 INFO [train.py:901] (0/4) Epoch 5, batch 6650, loss[loss=0.2709, simple_loss=0.3401, pruned_loss=0.1008, over 8537.00 frames. ], tot_loss[loss=0.2891, simple_loss=0.3509, pruned_loss=0.1137, over 1622616.36 frames. ], batch size: 31, lr: 1.45e-02, grad_scale: 8.0 +2023-02-06 01:31:15,369 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 +2023-02-06 01:31:29,865 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 5, batch 6700, loss[loss=0.2811, simple_loss=0.3504, pruned_loss=0.1059, over 8481.00 frames. ], tot_loss[loss=0.2878, simple_loss=0.3496, pruned_loss=0.113, over 1614677.29 frames. ], batch size: 29, lr: 1.45e-02, grad_scale: 8.0 +2023-02-06 01:32:08,107 INFO [train.py:901] (0/4) Epoch 5, batch 6750, loss[loss=0.2429, simple_loss=0.3232, pruned_loss=0.0813, over 8248.00 frames. ], tot_loss[loss=0.2875, simple_loss=0.3496, pruned_loss=0.1127, over 1613902.71 frames. ], batch size: 24, lr: 1.45e-02, grad_scale: 8.0 +2023-02-06 01:32:15,826 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39093.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:32:33,078 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 01:32:43,708 INFO [train.py:901] (0/4) Epoch 5, batch 6800, loss[loss=0.2848, simple_loss=0.3502, pruned_loss=0.1097, over 8103.00 frames. ], tot_loss[loss=0.2863, simple_loss=0.349, pruned_loss=0.1118, over 1617324.07 frames. ], batch size: 23, lr: 1.44e-02, grad_scale: 8.0 +2023-02-06 01:32:54,739 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-06 01:33:17,611 INFO [train.py:901] (0/4) Epoch 5, batch 6850, loss[loss=0.3284, simple_loss=0.3818, pruned_loss=0.1375, over 8198.00 frames. ], tot_loss[loss=0.2887, simple_loss=0.3506, pruned_loss=0.1134, over 1613815.18 frames. ], batch size: 23, lr: 1.44e-02, grad_scale: 8.0 +2023-02-06 01:33:19,090 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=39185.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:33:31,216 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39203.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:33:38,544 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2151, 1.3752, 4.0636, 1.7892, 2.1655, 4.7506, 4.7383, 4.1162], + device='cuda:0'), covar=tensor([0.1132, 0.1703, 0.0313, 0.2034, 0.0988, 0.0245, 0.0332, 0.0559], + device='cuda:0'), in_proj_covar=tensor([0.0235, 0.0266, 0.0224, 0.0262, 0.0224, 0.0199, 0.0231, 0.0272], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 01:33:43,525 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-06 01:33:48,726 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 5, batch 6900, loss[loss=0.3312, simple_loss=0.3717, pruned_loss=0.1453, over 7001.00 frames. ], tot_loss[loss=0.2888, simple_loss=0.3507, pruned_loss=0.1134, over 1610795.27 frames. ], batch size: 71, lr: 1.44e-02, grad_scale: 8.0 +2023-02-06 01:34:26,649 INFO [train.py:901] (0/4) Epoch 5, batch 6950, loss[loss=0.3896, simple_loss=0.4381, pruned_loss=0.1706, over 8579.00 frames. ], tot_loss[loss=0.2905, simple_loss=0.3524, pruned_loss=0.1143, over 1614561.65 frames. ], batch size: 31, lr: 1.44e-02, grad_scale: 8.0 +2023-02-06 01:34:38,143 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39300.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:34:39,400 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39302.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:34:49,519 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-06 01:34:58,305 INFO [optim.py:369] (0/4) 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 01:35:01,671 INFO [train.py:901] (0/4) Epoch 5, batch 7000, loss[loss=0.2679, simple_loss=0.33, pruned_loss=0.1029, over 7928.00 frames. ], tot_loss[loss=0.2898, simple_loss=0.3513, pruned_loss=0.1141, over 1610269.58 frames. ], batch size: 20, lr: 1.44e-02, grad_scale: 8.0 +2023-02-06 01:35:02,173 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 +2023-02-06 01:35:35,811 INFO [train.py:901] (0/4) Epoch 5, batch 7050, loss[loss=0.2745, simple_loss=0.3225, pruned_loss=0.1133, over 8083.00 frames. ], tot_loss[loss=0.2897, simple_loss=0.351, pruned_loss=0.1142, over 1612065.11 frames. ], batch size: 21, lr: 1.44e-02, grad_scale: 8.0 +2023-02-06 01:35:40,647 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8098, 2.7643, 1.7584, 4.2186, 1.8889, 1.5491, 2.4057, 3.0322], + device='cuda:0'), covar=tensor([0.2272, 0.1749, 0.3343, 0.0340, 0.2376, 0.3144, 0.2149, 0.1233], + device='cuda:0'), in_proj_covar=tensor([0.0276, 0.0254, 0.0289, 0.0229, 0.0248, 0.0282, 0.0287, 0.0257], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 01:36:06,894 INFO [optim.py:369] (0/4) 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 01:36:10,286 INFO [train.py:901] (0/4) Epoch 5, batch 7100, loss[loss=0.2728, simple_loss=0.3451, pruned_loss=0.1002, over 8038.00 frames. ], tot_loss[loss=0.2874, simple_loss=0.3493, pruned_loss=0.1127, over 1613557.58 frames. ], batch size: 22, lr: 1.44e-02, grad_scale: 8.0 +2023-02-06 01:36:44,807 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39481.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:36:46,011 INFO [train.py:901] (0/4) Epoch 5, batch 7150, loss[loss=0.2899, simple_loss=0.3606, pruned_loss=0.1096, over 8100.00 frames. ], tot_loss[loss=0.2858, simple_loss=0.3478, pruned_loss=0.1118, over 1608936.50 frames. ], batch size: 23, lr: 1.44e-02, grad_scale: 8.0 +2023-02-06 01:37:17,186 INFO [optim.py:369] (0/4) 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 01:37:20,758 INFO [train.py:901] (0/4) Epoch 5, batch 7200, loss[loss=0.2471, simple_loss=0.2999, pruned_loss=0.09717, over 7284.00 frames. ], tot_loss[loss=0.2867, simple_loss=0.3483, pruned_loss=0.1126, over 1608925.88 frames. ], batch size: 16, lr: 1.44e-02, grad_scale: 8.0 +2023-02-06 01:37:30,581 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=39547.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:37:37,481 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39556.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:37:51,898 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 +2023-02-06 01:37:55,143 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39581.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:37:56,305 INFO [train.py:901] (0/4) Epoch 5, batch 7250, loss[loss=0.3129, simple_loss=0.3756, pruned_loss=0.1251, over 8324.00 frames. ], tot_loss[loss=0.2844, simple_loss=0.3465, pruned_loss=0.1112, over 1607004.08 frames. ], batch size: 25, lr: 1.44e-02, grad_scale: 8.0 +2023-02-06 01:38:27,208 INFO [optim.py:369] (0/4) 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 01:38:30,508 INFO [train.py:901] (0/4) Epoch 5, batch 7300, loss[loss=0.3115, simple_loss=0.3777, pruned_loss=0.1227, over 7969.00 frames. ], tot_loss[loss=0.2858, simple_loss=0.3474, pruned_loss=0.1121, over 1606903.06 frames. ], batch size: 21, lr: 1.44e-02, grad_scale: 8.0 +2023-02-06 01:38:39,559 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=39646.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:38:44,000 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-02-06 01:38:47,105 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39657.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 01:38:50,501 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39662.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:39:05,555 INFO [train.py:901] (0/4) Epoch 5, batch 7350, loss[loss=0.2789, simple_loss=0.3506, pruned_loss=0.1036, over 8450.00 frames. ], tot_loss[loss=0.2851, simple_loss=0.3471, pruned_loss=0.1115, over 1606969.25 frames. ], batch size: 27, lr: 1.43e-02, grad_scale: 8.0 +2023-02-06 01:39:07,072 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.5453, 1.7903, 3.1392, 2.6698, 2.7206, 1.9169, 1.4536, 1.3726], + device='cuda:0'), covar=tensor([0.1751, 0.2503, 0.0476, 0.1076, 0.0962, 0.1155, 0.1276, 0.2177], + device='cuda:0'), in_proj_covar=tensor([0.0755, 0.0687, 0.0584, 0.0671, 0.0773, 0.0644, 0.0609, 0.0633], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 01:39:33,436 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-06 01:39:36,153 INFO [optim.py:369] (0/4) 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 01:39:39,621 INFO [train.py:901] (0/4) Epoch 5, batch 7400, loss[loss=0.2671, simple_loss=0.3215, pruned_loss=0.1064, over 7210.00 frames. ], tot_loss[loss=0.2852, simple_loss=0.3475, pruned_loss=0.1114, over 1613349.51 frames. ], batch size: 16, lr: 1.43e-02, grad_scale: 8.0 +2023-02-06 01:39:52,996 WARNING [train.py:1067] (0/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-06 01:39:59,142 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39761.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:40:13,744 INFO [train.py:901] (0/4) Epoch 5, batch 7450, loss[loss=0.2479, simple_loss=0.3235, pruned_loss=0.08615, over 8297.00 frames. ], tot_loss[loss=0.2881, simple_loss=0.3496, pruned_loss=0.1133, over 1612849.78 frames. ], batch size: 23, lr: 1.43e-02, grad_scale: 8.0 +2023-02-06 01:40:15,949 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39785.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:40:32,192 WARNING [train.py:1067] (0/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-06 01:40:33,685 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4671, 1.0375, 1.4656, 1.0976, 1.5650, 1.2789, 1.8750, 1.9653], + device='cuda:0'), covar=tensor([0.0683, 0.2141, 0.2953, 0.2027, 0.0690, 0.2546, 0.0920, 0.0660], + device='cuda:0'), in_proj_covar=tensor([0.0132, 0.0177, 0.0216, 0.0180, 0.0128, 0.0186, 0.0142, 0.0150], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0007, 0.0008, 0.0007, 0.0005, 0.0007, 0.0006, 0.0006], + device='cuda:0') +2023-02-06 01:40:43,630 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=39825.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:40:45,620 INFO [optim.py:369] (0/4) 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 01:40:48,852 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.23 vs. limit=5.0 +2023-02-06 01:40:49,055 INFO [train.py:901] (0/4) Epoch 5, batch 7500, loss[loss=0.2252, simple_loss=0.3037, pruned_loss=0.07338, over 8236.00 frames. ], tot_loss[loss=0.2861, simple_loss=0.348, pruned_loss=0.1121, over 1610977.45 frames. ], batch size: 22, lr: 1.43e-02, grad_scale: 8.0 +2023-02-06 01:40:49,193 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39833.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:41:22,987 INFO [train.py:901] (0/4) Epoch 5, batch 7550, loss[loss=0.2328, simple_loss=0.3127, pruned_loss=0.07646, over 8299.00 frames. ], tot_loss[loss=0.2847, simple_loss=0.3466, pruned_loss=0.1114, over 1610144.58 frames. ], batch size: 23, lr: 1.43e-02, grad_scale: 8.0 +2023-02-06 01:41:43,308 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1155, 2.7455, 3.2460, 1.3653, 3.2818, 2.0815, 1.5710, 1.8332], + device='cuda:0'), covar=tensor([0.0259, 0.0115, 0.0081, 0.0244, 0.0177, 0.0321, 0.0313, 0.0203], + device='cuda:0'), in_proj_covar=tensor([0.0311, 0.0222, 0.0191, 0.0273, 0.0218, 0.0364, 0.0284, 0.0261], + device='cuda:0'), out_proj_covar=tensor([1.0851e-04, 7.5473e-05, 6.4940e-05, 9.3373e-05, 7.6239e-05, 1.3584e-04, + 9.9626e-05, 8.9550e-05], device='cuda:0') +2023-02-06 01:41:48,005 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) Epoch 5, batch 7600, loss[loss=0.3022, simple_loss=0.374, pruned_loss=0.1152, over 8339.00 frames. ], tot_loss[loss=0.283, simple_loss=0.3452, pruned_loss=0.1104, over 1609177.41 frames. ], batch size: 25, lr: 1.43e-02, grad_scale: 8.0 +2023-02-06 01:42:02,463 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39940.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:42:04,534 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39943.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:42:07,192 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.9687, 1.5550, 5.9435, 2.1461, 5.2383, 5.0767, 5.5265, 5.4532], + device='cuda:0'), covar=tensor([0.0338, 0.3813, 0.0250, 0.2505, 0.0942, 0.0499, 0.0346, 0.0418], + device='cuda:0'), in_proj_covar=tensor([0.0335, 0.0488, 0.0430, 0.0424, 0.0492, 0.0412, 0.0398, 0.0452], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 01:42:32,755 INFO [train.py:901] (0/4) Epoch 5, batch 7650, loss[loss=0.2612, simple_loss=0.3428, pruned_loss=0.08985, over 8480.00 frames. ], tot_loss[loss=0.2837, simple_loss=0.3458, pruned_loss=0.1108, over 1609997.83 frames. ], batch size: 29, lr: 1.43e-02, grad_scale: 16.0 +2023-02-06 01:42:44,239 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-40000.pt +2023-02-06 01:42:45,878 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=40001.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 01:42:54,832 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9377, 1.5833, 3.3962, 1.3987, 2.1247, 3.8754, 3.8018, 3.2629], + device='cuda:0'), covar=tensor([0.1037, 0.1344, 0.0328, 0.1832, 0.0802, 0.0213, 0.0373, 0.0588], + device='cuda:0'), in_proj_covar=tensor([0.0236, 0.0263, 0.0223, 0.0263, 0.0222, 0.0201, 0.0235, 0.0276], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 01:42:57,013 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40017.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:43:05,596 INFO [optim.py:369] (0/4) 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 01:43:08,884 INFO [train.py:901] (0/4) Epoch 5, batch 7700, loss[loss=0.3285, simple_loss=0.3718, pruned_loss=0.1426, over 8336.00 frames. ], tot_loss[loss=0.2841, simple_loss=0.3461, pruned_loss=0.1111, over 1607260.69 frames. ], batch size: 26, lr: 1.43e-02, grad_scale: 16.0 +2023-02-06 01:43:15,453 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40042.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:43:44,285 INFO [train.py:901] (0/4) Epoch 5, batch 7750, loss[loss=0.2579, simple_loss=0.3268, pruned_loss=0.09446, over 8031.00 frames. ], tot_loss[loss=0.2848, simple_loss=0.3467, pruned_loss=0.1115, over 1606652.45 frames. ], batch size: 22, lr: 1.43e-02, grad_scale: 16.0 +2023-02-06 01:43:44,301 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-06 01:43:47,496 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.15 vs. limit=5.0 +2023-02-06 01:44:07,444 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40116.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 01:44:15,427 INFO [optim.py:369] (0/4) 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 01:44:16,242 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=40129.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:44:18,946 INFO [train.py:901] (0/4) Epoch 5, batch 7800, loss[loss=0.2494, simple_loss=0.3312, pruned_loss=0.08381, over 8287.00 frames. ], tot_loss[loss=0.2842, simple_loss=0.3458, pruned_loss=0.1113, over 1606891.16 frames. ], batch size: 23, lr: 1.43e-02, grad_scale: 16.0 +2023-02-06 01:44:38,906 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4870, 1.9798, 1.9327, 0.8727, 2.0363, 1.3182, 0.3968, 1.6236], + device='cuda:0'), covar=tensor([0.0152, 0.0090, 0.0077, 0.0165, 0.0098, 0.0313, 0.0282, 0.0091], + device='cuda:0'), in_proj_covar=tensor([0.0312, 0.0225, 0.0192, 0.0275, 0.0220, 0.0367, 0.0291, 0.0265], + device='cuda:0'), out_proj_covar=tensor([1.0823e-04, 7.6175e-05, 6.4895e-05, 9.3683e-05, 7.6646e-05, 1.3668e-04, + 1.0178e-04, 9.1007e-05], device='cuda:0') +2023-02-06 01:44:50,158 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=40177.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:44:54,273 INFO [train.py:901] (0/4) Epoch 5, batch 7850, loss[loss=0.2509, simple_loss=0.3178, pruned_loss=0.09196, over 8247.00 frames. ], tot_loss[loss=0.2858, simple_loss=0.347, pruned_loss=0.1123, over 1611884.50 frames. ], batch size: 24, lr: 1.43e-02, grad_scale: 16.0 +2023-02-06 01:45:03,223 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40196.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:45:14,682 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40213.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:45:20,135 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40221.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:45:24,750 INFO [optim.py:369] (0/4) 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 01:45:28,290 INFO [train.py:901] (0/4) Epoch 5, batch 7900, loss[loss=0.3973, simple_loss=0.4292, pruned_loss=0.1827, over 7049.00 frames. ], tot_loss[loss=0.2853, simple_loss=0.3472, pruned_loss=0.1117, over 1609958.52 frames. ], batch size: 71, lr: 1.42e-02, grad_scale: 16.0 +2023-02-06 01:45:30,521 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40236.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:45:35,953 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40244.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:46:02,038 INFO [train.py:901] (0/4) Epoch 5, batch 7950, loss[loss=0.2317, simple_loss=0.302, pruned_loss=0.08071, over 7684.00 frames. ], tot_loss[loss=0.2843, simple_loss=0.3459, pruned_loss=0.1113, over 1607069.51 frames. ], batch size: 18, lr: 1.42e-02, grad_scale: 16.0 +2023-02-06 01:46:07,572 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40290.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:46:09,034 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40292.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:46:33,074 INFO [optim.py:369] (0/4) 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 01:46:36,404 INFO [train.py:901] (0/4) Epoch 5, batch 8000, loss[loss=0.3276, simple_loss=0.3869, pruned_loss=0.1341, over 8338.00 frames. ], tot_loss[loss=0.2862, simple_loss=0.3478, pruned_loss=0.1123, over 1613490.56 frames. ], batch size: 26, lr: 1.42e-02, grad_scale: 16.0 +2023-02-06 01:46:46,588 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40348.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:47:03,164 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40372.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 01:47:10,480 INFO [train.py:901] (0/4) Epoch 5, batch 8050, loss[loss=0.2133, simple_loss=0.2724, pruned_loss=0.07705, over 7539.00 frames. ], tot_loss[loss=0.2853, simple_loss=0.3468, pruned_loss=0.1119, over 1606116.83 frames. ], batch size: 18, lr: 1.42e-02, grad_scale: 8.0 +2023-02-06 01:47:20,239 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40397.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 01:47:23,690 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3556, 2.3314, 2.0770, 1.2816, 2.0408, 1.9752, 2.1481, 1.8854], + device='cuda:0'), covar=tensor([0.1088, 0.0819, 0.1038, 0.3436, 0.0967, 0.1117, 0.1275, 0.0873], + device='cuda:0'), in_proj_covar=tensor([0.0404, 0.0284, 0.0323, 0.0412, 0.0312, 0.0277, 0.0303, 0.0250], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 01:47:33,786 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/epoch-5.pt +2023-02-06 01:47:44,680 WARNING [train.py:1067] (0/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-06 01:47:48,104 INFO [train.py:901] (0/4) Epoch 6, batch 0, loss[loss=0.2632, simple_loss=0.3334, pruned_loss=0.09648, over 8242.00 frames. ], tot_loss[loss=0.2632, simple_loss=0.3334, pruned_loss=0.09648, over 8242.00 frames. ], batch size: 22, lr: 1.33e-02, grad_scale: 8.0 +2023-02-06 01:47:48,104 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 01:47:59,053 INFO [train.py:935] (0/4) Epoch 6, validation: loss=0.2203, simple_loss=0.3165, pruned_loss=0.06206, over 944034.00 frames. +2023-02-06 01:47:59,054 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6689MB +2023-02-06 01:48:07,794 INFO [optim.py:369] (0/4) 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 01:48:13,423 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-06 01:48:14,983 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3225, 1.4860, 1.2407, 1.9851, 0.8418, 1.1170, 1.2594, 1.4996], + device='cuda:0'), covar=tensor([0.1397, 0.1209, 0.1948, 0.0742, 0.1569, 0.2299, 0.1240, 0.1089], + device='cuda:0'), in_proj_covar=tensor([0.0278, 0.0256, 0.0287, 0.0226, 0.0250, 0.0281, 0.0282, 0.0259], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 01:48:17,292 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 +2023-02-06 01:48:34,118 INFO [train.py:901] (0/4) Epoch 6, batch 50, loss[loss=0.2515, simple_loss=0.3118, pruned_loss=0.09554, over 7794.00 frames. ], tot_loss[loss=0.2876, simple_loss=0.3507, pruned_loss=0.1122, over 369705.46 frames. ], batch size: 19, lr: 1.33e-02, grad_scale: 8.0 +2023-02-06 01:48:48,504 WARNING [train.py:1067] (0/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-06 01:48:57,358 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40500.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:49:04,769 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40510.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:49:08,809 INFO [train.py:901] (0/4) Epoch 6, batch 100, loss[loss=0.3516, simple_loss=0.3967, pruned_loss=0.1532, over 8342.00 frames. ], tot_loss[loss=0.2902, simple_loss=0.3529, pruned_loss=0.1138, over 651350.97 frames. ], batch size: 26, lr: 1.33e-02, grad_scale: 8.0 +2023-02-06 01:49:13,092 WARNING [train.py:1067] (0/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-06 01:49:15,329 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40525.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:49:17,934 INFO [optim.py:369] (0/4) 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 01:49:27,173 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.0064, 1.4043, 4.2457, 1.5328, 3.6707, 3.5632, 3.8114, 3.6652], + device='cuda:0'), covar=tensor([0.0480, 0.3145, 0.0357, 0.2332, 0.1018, 0.0612, 0.0456, 0.0567], + device='cuda:0'), in_proj_covar=tensor([0.0344, 0.0490, 0.0436, 0.0431, 0.0499, 0.0418, 0.0407, 0.0458], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 01:49:31,507 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40548.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:49:37,659 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=40557.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:49:39,059 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.6968, 5.7743, 5.0029, 2.0083, 5.0483, 5.3438, 5.3048, 4.9598], + device='cuda:0'), covar=tensor([0.0612, 0.0430, 0.0776, 0.4748, 0.0672, 0.0756, 0.1116, 0.0670], + device='cuda:0'), in_proj_covar=tensor([0.0403, 0.0288, 0.0324, 0.0414, 0.0316, 0.0283, 0.0304, 0.0254], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 01:49:44,253 INFO [train.py:901] (0/4) Epoch 6, batch 150, loss[loss=0.2317, simple_loss=0.3204, pruned_loss=0.07151, over 8332.00 frames. ], tot_loss[loss=0.2919, simple_loss=0.3546, pruned_loss=0.1146, over 868955.13 frames. ], batch size: 25, lr: 1.32e-02, grad_scale: 8.0 +2023-02-06 01:49:49,733 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40573.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:49:54,327 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=40580.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:50:04,510 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4237, 1.5942, 1.3477, 1.8680, 0.9184, 1.1951, 1.3318, 1.6129], + device='cuda:0'), covar=tensor([0.1165, 0.0971, 0.1558, 0.0700, 0.1394, 0.2115, 0.1237, 0.0827], + device='cuda:0'), in_proj_covar=tensor([0.0273, 0.0254, 0.0281, 0.0227, 0.0247, 0.0280, 0.0279, 0.0257], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 01:50:19,258 INFO [train.py:901] (0/4) Epoch 6, batch 200, loss[loss=0.2827, simple_loss=0.3517, pruned_loss=0.1069, over 8742.00 frames. ], tot_loss[loss=0.2856, simple_loss=0.3491, pruned_loss=0.111, over 1041131.89 frames. ], batch size: 30, lr: 1.32e-02, grad_scale: 8.0 +2023-02-06 01:50:28,762 INFO [optim.py:369] (0/4) 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 01:50:32,275 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=40634.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:50:54,066 INFO [train.py:901] (0/4) Epoch 6, batch 250, loss[loss=0.2988, simple_loss=0.3688, pruned_loss=0.1144, over 8335.00 frames. ], tot_loss[loss=0.2846, simple_loss=0.3482, pruned_loss=0.1105, over 1166216.85 frames. ], batch size: 26, lr: 1.32e-02, grad_scale: 8.0 +2023-02-06 01:50:56,282 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40669.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:50:58,382 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40672.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:50:59,877 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.02 vs. limit=5.0 +2023-02-06 01:51:04,085 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-06 01:51:12,275 WARNING [train.py:1067] (0/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-06 01:51:13,021 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=40692.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:51:15,100 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40695.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:51:29,252 INFO [train.py:901] (0/4) Epoch 6, batch 300, loss[loss=0.2247, simple_loss=0.2798, pruned_loss=0.08482, over 7516.00 frames. ], tot_loss[loss=0.2843, simple_loss=0.3476, pruned_loss=0.1105, over 1265930.38 frames. ], batch size: 18, lr: 1.32e-02, grad_scale: 8.0 +2023-02-06 01:51:38,588 INFO [optim.py:369] (0/4) 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 01:51:52,720 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 6, batch 350, loss[loss=0.2468, simple_loss=0.3, pruned_loss=0.09678, over 7689.00 frames. ], tot_loss[loss=0.2834, simple_loss=0.3467, pruned_loss=0.1101, over 1341678.59 frames. ], batch size: 18, lr: 1.32e-02, grad_scale: 8.0 +2023-02-06 01:52:17,247 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5507, 1.9432, 1.9601, 0.9001, 2.0161, 1.4210, 0.3587, 1.6846], + device='cuda:0'), covar=tensor([0.0181, 0.0115, 0.0095, 0.0165, 0.0130, 0.0334, 0.0296, 0.0094], + device='cuda:0'), in_proj_covar=tensor([0.0324, 0.0234, 0.0195, 0.0282, 0.0228, 0.0374, 0.0300, 0.0274], + device='cuda:0'), out_proj_covar=tensor([1.1224e-04, 7.8652e-05, 6.5579e-05, 9.5827e-05, 7.8811e-05, 1.3856e-04, + 1.0426e-04, 9.3391e-05], device='cuda:0') +2023-02-06 01:52:32,424 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40807.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:52:38,236 INFO [train.py:901] (0/4) Epoch 6, batch 400, loss[loss=0.2342, simple_loss=0.3039, pruned_loss=0.0822, over 7791.00 frames. ], tot_loss[loss=0.2839, simple_loss=0.3471, pruned_loss=0.1104, over 1405916.21 frames. ], batch size: 19, lr: 1.32e-02, grad_scale: 8.0 +2023-02-06 01:52:46,905 INFO [optim.py:369] (0/4) 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 01:53:04,898 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=40854.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:53:12,732 INFO [train.py:901] (0/4) Epoch 6, batch 450, loss[loss=0.236, simple_loss=0.3103, pruned_loss=0.0808, over 8085.00 frames. ], tot_loss[loss=0.2838, simple_loss=0.3471, pruned_loss=0.1103, over 1451608.41 frames. ], batch size: 21, lr: 1.32e-02, grad_scale: 8.0 +2023-02-06 01:53:24,456 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40883.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:53:47,833 INFO [train.py:901] (0/4) Epoch 6, batch 500, loss[loss=0.2995, simple_loss=0.3558, pruned_loss=0.1216, over 8627.00 frames. ], tot_loss[loss=0.2824, simple_loss=0.3463, pruned_loss=0.1092, over 1492191.42 frames. ], batch size: 34, lr: 1.32e-02, grad_scale: 8.0 +2023-02-06 01:53:49,255 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) 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 01:54:12,294 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40951.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:54:13,635 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40953.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:54:22,945 INFO [train.py:901] (0/4) Epoch 6, batch 550, loss[loss=0.2453, simple_loss=0.3088, pruned_loss=0.09088, over 7693.00 frames. ], tot_loss[loss=0.2792, simple_loss=0.3439, pruned_loss=0.1072, over 1522784.68 frames. ], batch size: 18, lr: 1.32e-02, grad_scale: 8.0 +2023-02-06 01:54:25,204 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40969.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:54:30,610 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40976.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:54:49,826 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41005.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:54:55,069 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=41013.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:54:56,475 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41015.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:54:57,037 INFO [train.py:901] (0/4) Epoch 6, batch 600, loss[loss=0.3228, simple_loss=0.3714, pruned_loss=0.1371, over 8548.00 frames. ], tot_loss[loss=0.2788, simple_loss=0.3434, pruned_loss=0.1071, over 1544432.91 frames. ], batch size: 31, lr: 1.32e-02, grad_scale: 8.0 +2023-02-06 01:55:06,096 INFO [optim.py:369] (0/4) 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 01:55:06,982 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41030.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:55:09,481 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-06 01:55:29,303 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41063.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:55:29,383 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41063.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:55:31,087 INFO [train.py:901] (0/4) Epoch 6, batch 650, loss[loss=0.2793, simple_loss=0.3515, pruned_loss=0.1036, over 8257.00 frames. ], tot_loss[loss=0.2812, simple_loss=0.3452, pruned_loss=0.1086, over 1562462.51 frames. ], batch size: 24, lr: 1.32e-02, grad_scale: 8.0 +2023-02-06 01:55:43,234 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.11 vs. limit=5.0 +2023-02-06 01:55:46,207 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41088.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:56:05,824 INFO [train.py:901] (0/4) Epoch 6, batch 700, loss[loss=0.2509, simple_loss=0.3152, pruned_loss=0.09332, over 7773.00 frames. ], tot_loss[loss=0.2829, simple_loss=0.3462, pruned_loss=0.1098, over 1574631.49 frames. ], batch size: 19, lr: 1.32e-02, grad_scale: 8.0 +2023-02-06 01:56:08,623 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41120.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:56:14,230 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41128.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:56:14,741 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 6, batch 750, loss[loss=0.3432, simple_loss=0.388, pruned_loss=0.1492, over 8478.00 frames. ], tot_loss[loss=0.2836, simple_loss=0.3469, pruned_loss=0.1102, over 1585084.56 frames. ], batch size: 29, lr: 1.31e-02, grad_scale: 8.0 +2023-02-06 01:56:52,792 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-06 01:57:00,960 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 6, batch 800, loss[loss=0.2118, simple_loss=0.2949, pruned_loss=0.06435, over 8190.00 frames. ], tot_loss[loss=0.2836, simple_loss=0.3473, pruned_loss=0.1099, over 1598082.01 frames. ], batch size: 23, lr: 1.31e-02, grad_scale: 8.0 +2023-02-06 01:57:21,525 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41225.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:57:22,746 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=41227.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:57:24,006 INFO [optim.py:369] (0/4) 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 01:57:38,348 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41250.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:57:40,899 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.7462, 4.6994, 4.1348, 1.7778, 4.1295, 4.1648, 4.2970, 3.7573], + device='cuda:0'), covar=tensor([0.0642, 0.0504, 0.0841, 0.4648, 0.0693, 0.0799, 0.1302, 0.0653], + device='cuda:0'), in_proj_covar=tensor([0.0404, 0.0286, 0.0319, 0.0407, 0.0310, 0.0280, 0.0301, 0.0256], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 01:57:46,851 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=41262.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:57:49,530 INFO [train.py:901] (0/4) Epoch 6, batch 850, loss[loss=0.2537, simple_loss=0.3339, pruned_loss=0.0868, over 8038.00 frames. ], tot_loss[loss=0.2834, simple_loss=0.3468, pruned_loss=0.11, over 1601819.87 frames. ], batch size: 22, lr: 1.31e-02, grad_scale: 8.0 +2023-02-06 01:57:59,689 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.96 vs. limit=2.0 +2023-02-06 01:58:08,939 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3456, 2.0507, 3.1063, 2.5114, 2.5568, 2.0269, 1.5590, 1.3440], + device='cuda:0'), covar=tensor([0.2111, 0.2038, 0.0492, 0.1175, 0.1037, 0.1193, 0.1193, 0.2226], + device='cuda:0'), in_proj_covar=tensor([0.0770, 0.0702, 0.0601, 0.0695, 0.0797, 0.0660, 0.0624, 0.0650], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 01:58:23,892 INFO [train.py:901] (0/4) Epoch 6, batch 900, loss[loss=0.2911, simple_loss=0.3569, pruned_loss=0.1126, over 8498.00 frames. ], tot_loss[loss=0.2817, simple_loss=0.3458, pruned_loss=0.1088, over 1608520.88 frames. ], batch size: 26, lr: 1.31e-02, grad_scale: 8.0 +2023-02-06 01:58:28,819 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.1782, 1.6075, 5.2624, 2.0909, 4.6217, 4.5509, 4.8685, 4.8638], + device='cuda:0'), covar=tensor([0.0377, 0.3549, 0.0373, 0.2419, 0.0970, 0.0614, 0.0407, 0.0395], + device='cuda:0'), in_proj_covar=tensor([0.0355, 0.0510, 0.0449, 0.0441, 0.0511, 0.0434, 0.0420, 0.0474], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 01:58:33,478 INFO [optim.py:369] (0/4) 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 01:58:42,519 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41342.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:58:53,863 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=41359.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:58:58,557 INFO [train.py:901] (0/4) Epoch 6, batch 950, loss[loss=0.2345, simple_loss=0.3057, pruned_loss=0.08165, over 7538.00 frames. ], tot_loss[loss=0.2813, simple_loss=0.3453, pruned_loss=0.1087, over 1609500.06 frames. ], batch size: 18, lr: 1.31e-02, grad_scale: 8.0 +2023-02-06 01:59:06,373 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41377.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:59:11,053 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41384.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:59:21,436 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-06 01:59:26,984 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=41407.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 01:59:28,455 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 6, batch 1000, loss[loss=0.2594, simple_loss=0.3281, pruned_loss=0.09532, over 8239.00 frames. ], tot_loss[loss=0.2801, simple_loss=0.3442, pruned_loss=0.1081, over 1612013.98 frames. ], batch size: 22, lr: 1.31e-02, grad_scale: 8.0 +2023-02-06 01:59:33,844 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6972, 1.9711, 1.6390, 2.3806, 1.2315, 1.3292, 1.7219, 2.0560], + device='cuda:0'), covar=tensor([0.1081, 0.1083, 0.1510, 0.0643, 0.1487, 0.2104, 0.1222, 0.0741], + device='cuda:0'), in_proj_covar=tensor([0.0277, 0.0253, 0.0284, 0.0227, 0.0250, 0.0282, 0.0287, 0.0261], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 01:59:41,362 INFO [optim.py:369] (0/4) 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 01:59:55,293 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-06 02:00:06,232 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 6, batch 1050, loss[loss=0.2292, simple_loss=0.2974, pruned_loss=0.08049, over 7680.00 frames. ], tot_loss[loss=0.2805, simple_loss=0.3447, pruned_loss=0.1082, over 1616931.01 frames. ], batch size: 18, lr: 1.31e-02, grad_scale: 8.0 +2023-02-06 02:00:08,220 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-06 02:00:13,132 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41474.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:00:19,810 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.1115, 1.5881, 1.5195, 1.3075, 1.4614, 1.6444, 2.1275, 2.0691], + device='cuda:0'), covar=tensor([0.0619, 0.1821, 0.2734, 0.2018, 0.0760, 0.2179, 0.0849, 0.0682], + device='cuda:0'), in_proj_covar=tensor([0.0131, 0.0176, 0.0219, 0.0183, 0.0130, 0.0188, 0.0140, 0.0150], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0007, 0.0008, 0.0007, 0.0005, 0.0007, 0.0006, 0.0006], + device='cuda:0') +2023-02-06 02:00:42,284 INFO [train.py:901] (0/4) Epoch 6, batch 1100, loss[loss=0.2401, simple_loss=0.3255, pruned_loss=0.07737, over 8029.00 frames. ], tot_loss[loss=0.28, simple_loss=0.3442, pruned_loss=0.1079, over 1616731.58 frames. ], batch size: 22, lr: 1.31e-02, grad_scale: 8.0 +2023-02-06 02:00:46,701 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) Epoch 6, batch 1150, loss[loss=0.2666, simple_loss=0.3312, pruned_loss=0.101, over 8034.00 frames. ], tot_loss[loss=0.2797, simple_loss=0.3438, pruned_loss=0.1077, over 1615185.08 frames. ], batch size: 22, lr: 1.31e-02, grad_scale: 8.0 +2023-02-06 02:01:18,801 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-06 02:01:25,439 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41579.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:01:38,055 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41598.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:01:50,378 INFO [train.py:901] (0/4) Epoch 6, batch 1200, loss[loss=0.2501, simple_loss=0.3242, pruned_loss=0.08797, over 8511.00 frames. ], tot_loss[loss=0.2797, simple_loss=0.3437, pruned_loss=0.1078, over 1616054.06 frames. ], batch size: 26, lr: 1.31e-02, grad_scale: 8.0 +2023-02-06 02:01:55,784 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41623.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:01:58,418 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5689, 1.8577, 2.0709, 1.6986, 1.0311, 2.1005, 0.2934, 1.2221], + device='cuda:0'), covar=tensor([0.2160, 0.1630, 0.0685, 0.1598, 0.5111, 0.0607, 0.4223, 0.1976], + device='cuda:0'), in_proj_covar=tensor([0.0129, 0.0129, 0.0080, 0.0173, 0.0209, 0.0081, 0.0145, 0.0133], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 02:02:00,249 INFO [optim.py:369] (0/4) 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 02:02:03,207 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41658.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:02:24,691 INFO [train.py:901] (0/4) Epoch 6, batch 1250, loss[loss=0.3735, simple_loss=0.3982, pruned_loss=0.1744, over 6745.00 frames. ], tot_loss[loss=0.2798, simple_loss=0.3435, pruned_loss=0.1081, over 1615773.61 frames. ], batch size: 72, lr: 1.31e-02, grad_scale: 8.0 +2023-02-06 02:02:29,662 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 6, batch 1300, loss[loss=0.2071, simple_loss=0.2722, pruned_loss=0.07099, over 7704.00 frames. ], tot_loss[loss=0.2783, simple_loss=0.3425, pruned_loss=0.107, over 1618146.11 frames. ], batch size: 18, lr: 1.31e-02, grad_scale: 8.0 +2023-02-06 02:03:08,608 INFO [optim.py:369] (0/4) 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 02:03:09,507 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41730.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:03:27,418 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41755.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:03:34,647 INFO [train.py:901] (0/4) Epoch 6, batch 1350, loss[loss=0.2497, simple_loss=0.3113, pruned_loss=0.09404, over 7808.00 frames. ], tot_loss[loss=0.2791, simple_loss=0.3432, pruned_loss=0.1075, over 1617453.79 frames. ], batch size: 20, lr: 1.31e-02, grad_scale: 8.0 +2023-02-06 02:03:42,954 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41778.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:04:00,405 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41803.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:04:09,625 INFO [train.py:901] (0/4) Epoch 6, batch 1400, loss[loss=0.2454, simple_loss=0.2984, pruned_loss=0.0962, over 7535.00 frames. ], tot_loss[loss=0.2783, simple_loss=0.3426, pruned_loss=0.107, over 1621581.02 frames. ], batch size: 18, lr: 1.30e-02, grad_scale: 8.0 +2023-02-06 02:04:18,120 INFO [optim.py:369] (0/4) 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 02:04:22,369 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41835.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:04:39,964 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41860.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:04:44,589 INFO [train.py:901] (0/4) Epoch 6, batch 1450, loss[loss=0.374, simple_loss=0.3985, pruned_loss=0.1748, over 6532.00 frames. ], tot_loss[loss=0.2785, simple_loss=0.3431, pruned_loss=0.107, over 1620639.77 frames. ], batch size: 71, lr: 1.30e-02, grad_scale: 8.0 +2023-02-06 02:04:47,875 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-06 02:05:10,758 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.1794, 1.2558, 4.2256, 1.6158, 3.7325, 3.5545, 3.8739, 3.7989], + device='cuda:0'), covar=tensor([0.0405, 0.3851, 0.0449, 0.2480, 0.1141, 0.0755, 0.0443, 0.0499], + device='cuda:0'), in_proj_covar=tensor([0.0347, 0.0500, 0.0443, 0.0436, 0.0500, 0.0419, 0.0410, 0.0463], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 02:05:18,593 INFO [train.py:901] (0/4) Epoch 6, batch 1500, loss[loss=0.3303, simple_loss=0.3946, pruned_loss=0.133, over 8030.00 frames. ], tot_loss[loss=0.2806, simple_loss=0.3446, pruned_loss=0.1083, over 1617204.39 frames. ], batch size: 22, lr: 1.30e-02, grad_scale: 8.0 +2023-02-06 02:05:24,645 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41924.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:05:27,836 INFO [optim.py:369] (0/4) 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 02:05:53,224 INFO [train.py:901] (0/4) Epoch 6, batch 1550, loss[loss=0.3163, simple_loss=0.375, pruned_loss=0.1288, over 8281.00 frames. ], tot_loss[loss=0.2813, simple_loss=0.3453, pruned_loss=0.1087, over 1617143.25 frames. ], batch size: 23, lr: 1.30e-02, grad_scale: 8.0 +2023-02-06 02:05:53,373 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41966.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 02:06:16,688 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-42000.pt +2023-02-06 02:06:28,441 INFO [train.py:901] (0/4) Epoch 6, batch 1600, loss[loss=0.2985, simple_loss=0.379, pruned_loss=0.109, over 8672.00 frames. ], tot_loss[loss=0.2796, simple_loss=0.3438, pruned_loss=0.1078, over 1616791.20 frames. ], batch size: 34, lr: 1.30e-02, grad_scale: 8.0 +2023-02-06 02:06:28,511 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 02:06:47,626 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6348, 1.9243, 2.1073, 1.6343, 0.9353, 2.2926, 0.2856, 1.0899], + device='cuda:0'), covar=tensor([0.3380, 0.1998, 0.1003, 0.2383, 0.6790, 0.0548, 0.5109, 0.2498], + device='cuda:0'), in_proj_covar=tensor([0.0132, 0.0132, 0.0082, 0.0181, 0.0216, 0.0082, 0.0147, 0.0136], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 02:06:50,468 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 +2023-02-06 02:07:03,786 INFO [train.py:901] (0/4) Epoch 6, batch 1650, loss[loss=0.2855, simple_loss=0.344, pruned_loss=0.1136, over 8523.00 frames. ], tot_loss[loss=0.2787, simple_loss=0.343, pruned_loss=0.1072, over 1619350.28 frames. ], batch size: 39, lr: 1.30e-02, grad_scale: 8.0 +2023-02-06 02:07:27,173 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5715, 3.6996, 2.2665, 2.3421, 2.9129, 1.7604, 2.2361, 2.7138], + device='cuda:0'), covar=tensor([0.1431, 0.0281, 0.0796, 0.0737, 0.0505, 0.1157, 0.0965, 0.0799], + device='cuda:0'), in_proj_covar=tensor([0.0341, 0.0233, 0.0310, 0.0293, 0.0313, 0.0307, 0.0334, 0.0302], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 02:07:39,088 INFO [train.py:901] (0/4) Epoch 6, batch 1700, loss[loss=0.3145, simple_loss=0.3681, pruned_loss=0.1305, over 8199.00 frames. ], tot_loss[loss=0.2771, simple_loss=0.3418, pruned_loss=0.1062, over 1615784.97 frames. ], batch size: 23, lr: 1.30e-02, grad_scale: 8.0 +2023-02-06 02:07:44,497 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 +2023-02-06 02:07:47,895 INFO [optim.py:369] (0/4) 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 02:07:49,305 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=42131.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:08:14,471 INFO [train.py:901] (0/4) Epoch 6, batch 1750, loss[loss=0.2436, simple_loss=0.3056, pruned_loss=0.09086, over 8145.00 frames. ], tot_loss[loss=0.276, simple_loss=0.3402, pruned_loss=0.1058, over 1615729.90 frames. ], batch size: 22, lr: 1.30e-02, grad_scale: 8.0 +2023-02-06 02:08:42,599 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.3382, 2.2236, 1.7866, 1.6799, 1.4386, 1.8775, 2.3619, 2.0836], + device='cuda:0'), covar=tensor([0.0454, 0.1055, 0.1612, 0.1315, 0.0643, 0.1484, 0.0629, 0.0527], + device='cuda:0'), in_proj_covar=tensor([0.0128, 0.0174, 0.0213, 0.0178, 0.0127, 0.0183, 0.0139, 0.0149], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0007, 0.0008, 0.0007, 0.0005, 0.0007, 0.0006, 0.0006], + device='cuda:0') +2023-02-06 02:08:44,743 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7408, 1.6507, 1.9681, 1.8743, 1.1122, 2.2936, 0.6116, 1.2073], + device='cuda:0'), covar=tensor([0.2859, 0.2993, 0.1202, 0.1671, 0.5956, 0.0550, 0.5144, 0.2907], + device='cuda:0'), in_proj_covar=tensor([0.0132, 0.0133, 0.0084, 0.0178, 0.0218, 0.0082, 0.0144, 0.0136], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 02:08:49,394 INFO [train.py:901] (0/4) Epoch 6, batch 1800, loss[loss=0.3048, simple_loss=0.3655, pruned_loss=0.1221, over 8562.00 frames. ], tot_loss[loss=0.2762, simple_loss=0.3405, pruned_loss=0.1059, over 1615151.71 frames. ], batch size: 39, lr: 1.30e-02, grad_scale: 8.0 +2023-02-06 02:08:59,177 INFO [optim.py:369] (0/4) 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 02:09:24,925 INFO [train.py:901] (0/4) Epoch 6, batch 1850, loss[loss=0.2625, simple_loss=0.3367, pruned_loss=0.09416, over 8580.00 frames. ], tot_loss[loss=0.2763, simple_loss=0.341, pruned_loss=0.1058, over 1616050.20 frames. ], batch size: 31, lr: 1.30e-02, grad_scale: 8.0 +2023-02-06 02:09:26,411 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=42268.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:09:47,994 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([6.0184, 1.6995, 6.0087, 2.0398, 5.3374, 5.1578, 5.7362, 5.6104], + device='cuda:0'), covar=tensor([0.0415, 0.3722, 0.0287, 0.2457, 0.0979, 0.0522, 0.0397, 0.0396], + device='cuda:0'), in_proj_covar=tensor([0.0355, 0.0510, 0.0445, 0.0442, 0.0505, 0.0424, 0.0415, 0.0468], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 02:09:55,345 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=42310.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 02:09:59,278 INFO [train.py:901] (0/4) Epoch 6, batch 1900, loss[loss=0.2574, simple_loss=0.3317, pruned_loss=0.09157, over 8345.00 frames. ], tot_loss[loss=0.2742, simple_loss=0.3391, pruned_loss=0.1047, over 1613625.31 frames. ], batch size: 26, lr: 1.30e-02, grad_scale: 8.0 +2023-02-06 02:10:08,776 INFO [optim.py:369] (0/4) 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 02:10:19,490 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2950, 1.7788, 2.9720, 2.3238, 2.4432, 1.8816, 1.4750, 1.2067], + device='cuda:0'), covar=tensor([0.2061, 0.2508, 0.0541, 0.1252, 0.1055, 0.1382, 0.1331, 0.2228], + device='cuda:0'), in_proj_covar=tensor([0.0775, 0.0710, 0.0618, 0.0708, 0.0804, 0.0662, 0.0620, 0.0655], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 02:10:23,931 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-06 02:10:34,060 INFO [train.py:901] (0/4) Epoch 6, batch 1950, loss[loss=0.2577, simple_loss=0.3346, pruned_loss=0.09041, over 8475.00 frames. ], tot_loss[loss=0.2745, simple_loss=0.3397, pruned_loss=0.1046, over 1614331.14 frames. ], batch size: 29, lr: 1.30e-02, grad_scale: 16.0 +2023-02-06 02:10:36,640 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-06 02:10:37,662 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-06 02:10:46,201 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=42383.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:10:48,975 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42387.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:10:56,190 WARNING [train.py:1067] (0/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-06 02:10:56,312 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42397.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:11:06,394 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42412.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:11:08,998 INFO [train.py:901] (0/4) Epoch 6, batch 2000, loss[loss=0.3458, simple_loss=0.3865, pruned_loss=0.1525, over 8471.00 frames. ], tot_loss[loss=0.2748, simple_loss=0.3398, pruned_loss=0.1049, over 1616522.53 frames. ], batch size: 27, lr: 1.30e-02, grad_scale: 16.0 +2023-02-06 02:11:15,164 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=42425.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 02:11:18,300 INFO [optim.py:369] (0/4) 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 02:11:43,879 INFO [train.py:901] (0/4) Epoch 6, batch 2050, loss[loss=0.2163, simple_loss=0.2935, pruned_loss=0.06959, over 7537.00 frames. ], tot_loss[loss=0.2737, simple_loss=0.339, pruned_loss=0.1042, over 1616203.85 frames. ], batch size: 18, lr: 1.29e-02, grad_scale: 8.0 +2023-02-06 02:11:49,370 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.0939, 1.7786, 1.8374, 1.5532, 1.4287, 1.7197, 2.4566, 1.7469], + device='cuda:0'), covar=tensor([0.0496, 0.1219, 0.1687, 0.1425, 0.0683, 0.1530, 0.0664, 0.0602], + device='cuda:0'), in_proj_covar=tensor([0.0129, 0.0174, 0.0213, 0.0180, 0.0126, 0.0185, 0.0139, 0.0151], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0007, 0.0008, 0.0007, 0.0005, 0.0007, 0.0006, 0.0006], + device='cuda:0') +2023-02-06 02:12:17,264 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0355, 1.7325, 1.3564, 1.7873, 1.4355, 1.1618, 1.4529, 1.5795], + device='cuda:0'), covar=tensor([0.0749, 0.0334, 0.0854, 0.0390, 0.0500, 0.1026, 0.0608, 0.0515], + device='cuda:0'), in_proj_covar=tensor([0.0341, 0.0234, 0.0312, 0.0297, 0.0313, 0.0306, 0.0338, 0.0307], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 02:12:17,698 INFO [train.py:901] (0/4) Epoch 6, batch 2100, loss[loss=0.3192, simple_loss=0.3554, pruned_loss=0.1415, over 7817.00 frames. ], tot_loss[loss=0.2748, simple_loss=0.3403, pruned_loss=0.1046, over 1619852.57 frames. ], batch size: 20, lr: 1.29e-02, grad_scale: 8.0 +2023-02-06 02:12:23,857 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42524.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:12:27,687 INFO [optim.py:369] (0/4) 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 02:12:30,518 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.1990, 1.0311, 1.0964, 0.9809, 0.7462, 0.9773, 1.0200, 0.9083], + device='cuda:0'), covar=tensor([0.0491, 0.0894, 0.1257, 0.1032, 0.0500, 0.1074, 0.0565, 0.0441], + device='cuda:0'), in_proj_covar=tensor([0.0127, 0.0173, 0.0211, 0.0178, 0.0125, 0.0183, 0.0137, 0.0149], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0007, 0.0008, 0.0007, 0.0005, 0.0007, 0.0006, 0.0006], + device='cuda:0') +2023-02-06 02:12:52,377 INFO [train.py:901] (0/4) Epoch 6, batch 2150, loss[loss=0.2793, simple_loss=0.3283, pruned_loss=0.1152, over 7425.00 frames. ], tot_loss[loss=0.2772, simple_loss=0.3421, pruned_loss=0.1062, over 1619879.13 frames. ], batch size: 17, lr: 1.29e-02, grad_scale: 8.0 +2023-02-06 02:13:24,198 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 +2023-02-06 02:13:27,027 INFO [train.py:901] (0/4) Epoch 6, batch 2200, loss[loss=0.2427, simple_loss=0.3094, pruned_loss=0.08796, over 8241.00 frames. ], tot_loss[loss=0.2775, simple_loss=0.3424, pruned_loss=0.1063, over 1620569.66 frames. ], batch size: 22, lr: 1.29e-02, grad_scale: 8.0 +2023-02-06 02:13:36,143 INFO [optim.py:369] (0/4) 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] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42639.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:13:47,048 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5016, 1.7536, 2.0148, 0.7886, 2.1019, 1.3282, 0.5252, 1.7144], + device='cuda:0'), covar=tensor([0.0241, 0.0133, 0.0104, 0.0255, 0.0217, 0.0399, 0.0336, 0.0133], + device='cuda:0'), in_proj_covar=tensor([0.0331, 0.0239, 0.0211, 0.0297, 0.0234, 0.0388, 0.0305, 0.0277], + device='cuda:0'), out_proj_covar=tensor([1.1269e-04, 7.9082e-05, 6.9848e-05, 9.9620e-05, 7.9360e-05, 1.4094e-04, + 1.0432e-04, 9.3080e-05], device='cuda:0') +2023-02-06 02:13:59,699 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42664.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:14:00,825 INFO [train.py:901] (0/4) Epoch 6, batch 2250, loss[loss=0.2419, simple_loss=0.3264, pruned_loss=0.07865, over 8472.00 frames. ], tot_loss[loss=0.2766, simple_loss=0.3417, pruned_loss=0.1057, over 1619743.31 frames. ], batch size: 29, lr: 1.29e-02, grad_scale: 8.0 +2023-02-06 02:14:01,616 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42667.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:14:11,874 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42681.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 02:14:29,255 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42706.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 02:14:29,877 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.6565, 2.9919, 2.4327, 4.0258, 1.7981, 1.9057, 2.3112, 3.3331], + device='cuda:0'), covar=tensor([0.0863, 0.1041, 0.1138, 0.0263, 0.1592, 0.1875, 0.1450, 0.0819], + device='cuda:0'), in_proj_covar=tensor([0.0272, 0.0252, 0.0283, 0.0226, 0.0251, 0.0281, 0.0284, 0.0258], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 02:14:35,831 INFO [train.py:901] (0/4) Epoch 6, batch 2300, loss[loss=0.2352, simple_loss=0.3104, pruned_loss=0.07994, over 8137.00 frames. ], tot_loss[loss=0.2776, simple_loss=0.342, pruned_loss=0.1067, over 1618630.19 frames. ], batch size: 22, lr: 1.29e-02, grad_scale: 8.0 +2023-02-06 02:14:45,246 INFO [optim.py:369] (0/4) 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 02:14:53,234 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=42741.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:15:11,251 INFO [train.py:901] (0/4) Epoch 6, batch 2350, loss[loss=0.2946, simple_loss=0.3607, pruned_loss=0.1142, over 8351.00 frames. ], tot_loss[loss=0.2791, simple_loss=0.3425, pruned_loss=0.1078, over 1612400.27 frames. ], batch size: 24, lr: 1.29e-02, grad_scale: 8.0 +2023-02-06 02:15:46,588 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.88 vs. limit=5.0 +2023-02-06 02:15:46,903 INFO [train.py:901] (0/4) Epoch 6, batch 2400, loss[loss=0.2592, simple_loss=0.3286, pruned_loss=0.09491, over 8461.00 frames. ], tot_loss[loss=0.2803, simple_loss=0.3436, pruned_loss=0.1085, over 1618063.01 frames. ], batch size: 25, lr: 1.29e-02, grad_scale: 8.0 +2023-02-06 02:15:56,305 INFO [optim.py:369] (0/4) 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 02:16:08,777 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.09 vs. limit=5.0 +2023-02-06 02:16:14,310 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=42856.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:16:20,849 INFO [train.py:901] (0/4) Epoch 6, batch 2450, loss[loss=0.1987, simple_loss=0.2716, pruned_loss=0.06286, over 7196.00 frames. ], tot_loss[loss=0.2813, simple_loss=0.3445, pruned_loss=0.1091, over 1618204.23 frames. ], batch size: 16, lr: 1.29e-02, grad_scale: 8.0 +2023-02-06 02:16:22,280 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=42868.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:16:29,129 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42877.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:16:42,202 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42897.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:16:45,021 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.5754, 1.9898, 1.9859, 1.8964, 1.6498, 1.9927, 2.2053, 2.1214], + device='cuda:0'), covar=tensor([0.0663, 0.0976, 0.1348, 0.1121, 0.0597, 0.1196, 0.0767, 0.0449], + device='cuda:0'), in_proj_covar=tensor([0.0129, 0.0174, 0.0215, 0.0180, 0.0125, 0.0183, 0.0139, 0.0150], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0007, 0.0008, 0.0007, 0.0005, 0.0007, 0.0006, 0.0006], + device='cuda:0') +2023-02-06 02:16:54,609 INFO [train.py:901] (0/4) Epoch 6, batch 2500, loss[loss=0.3022, simple_loss=0.3666, pruned_loss=0.1189, over 8194.00 frames. ], tot_loss[loss=0.2812, simple_loss=0.3446, pruned_loss=0.1089, over 1615965.04 frames. ], batch size: 23, lr: 1.29e-02, grad_scale: 8.0 +2023-02-06 02:17:05,200 INFO [optim.py:369] (0/4) 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 02:17:29,434 INFO [train.py:901] (0/4) Epoch 6, batch 2550, loss[loss=0.3162, simple_loss=0.3528, pruned_loss=0.1398, over 7200.00 frames. ], tot_loss[loss=0.2818, simple_loss=0.345, pruned_loss=0.1093, over 1614860.09 frames. ], batch size: 16, lr: 1.29e-02, grad_scale: 8.0 +2023-02-06 02:17:41,648 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=42983.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:18:00,621 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9190, 1.6196, 3.3995, 1.4226, 2.1865, 3.8692, 3.7919, 3.2807], + device='cuda:0'), covar=tensor([0.1044, 0.1352, 0.0331, 0.1915, 0.0839, 0.0204, 0.0333, 0.0564], + device='cuda:0'), in_proj_covar=tensor([0.0239, 0.0267, 0.0225, 0.0264, 0.0236, 0.0209, 0.0248, 0.0279], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 02:18:01,213 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 6, batch 2600, loss[loss=0.2595, simple_loss=0.3338, pruned_loss=0.09264, over 8363.00 frames. ], tot_loss[loss=0.2826, simple_loss=0.3459, pruned_loss=0.1097, over 1619095.47 frames. ], batch size: 24, lr: 1.29e-02, grad_scale: 8.0 +2023-02-06 02:18:13,988 INFO [optim.py:369] (0/4) 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 02:18:39,586 INFO [train.py:901] (0/4) Epoch 6, batch 2650, loss[loss=0.2607, simple_loss=0.332, pruned_loss=0.09469, over 8290.00 frames. ], tot_loss[loss=0.2805, simple_loss=0.3439, pruned_loss=0.1085, over 1614809.38 frames. ], batch size: 23, lr: 1.29e-02, grad_scale: 8.0 +2023-02-06 02:18:47,158 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9290, 1.9425, 2.4125, 1.7376, 0.9884, 2.2847, 0.3862, 1.1534], + device='cuda:0'), covar=tensor([0.3423, 0.4111, 0.0850, 0.3180, 0.7648, 0.0783, 0.5793, 0.2702], + device='cuda:0'), in_proj_covar=tensor([0.0133, 0.0131, 0.0081, 0.0179, 0.0217, 0.0082, 0.0141, 0.0133], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 02:19:05,101 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0206, 0.9230, 1.0097, 0.9780, 0.6986, 1.1037, 0.0363, 0.6237], + device='cuda:0'), covar=tensor([0.2764, 0.2234, 0.0948, 0.1892, 0.5193, 0.0749, 0.4521, 0.2616], + device='cuda:0'), in_proj_covar=tensor([0.0134, 0.0133, 0.0083, 0.0182, 0.0219, 0.0084, 0.0144, 0.0136], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 02:19:11,128 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=43112.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:19:14,338 INFO [train.py:901] (0/4) Epoch 6, batch 2700, loss[loss=0.2761, simple_loss=0.3432, pruned_loss=0.1045, over 8541.00 frames. ], tot_loss[loss=0.2787, simple_loss=0.3421, pruned_loss=0.1076, over 1615302.11 frames. ], batch size: 39, lr: 1.29e-02, grad_scale: 8.0 +2023-02-06 02:19:15,193 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4635, 1.8339, 3.1043, 1.1276, 2.1951, 1.7641, 1.5253, 1.8672], + device='cuda:0'), covar=tensor([0.1597, 0.2008, 0.0678, 0.3515, 0.1476, 0.2644, 0.1531, 0.2276], + device='cuda:0'), in_proj_covar=tensor([0.0472, 0.0462, 0.0529, 0.0542, 0.0594, 0.0528, 0.0449, 0.0601], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], + device='cuda:0') +2023-02-06 02:19:20,968 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43126.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:19:23,436 INFO [optim.py:369] (0/4) 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 02:19:28,345 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43137.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:19:29,792 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 +2023-02-06 02:19:47,952 INFO [train.py:901] (0/4) Epoch 6, batch 2750, loss[loss=0.2474, simple_loss=0.3175, pruned_loss=0.08862, over 8354.00 frames. ], tot_loss[loss=0.2768, simple_loss=0.3408, pruned_loss=0.1064, over 1616808.14 frames. ], batch size: 24, lr: 1.28e-02, grad_scale: 8.0 +2023-02-06 02:19:57,318 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9309, 2.4721, 3.1015, 1.0556, 3.1539, 2.0049, 1.4369, 1.6874], + device='cuda:0'), covar=tensor([0.0340, 0.0164, 0.0127, 0.0303, 0.0186, 0.0331, 0.0354, 0.0222], + device='cuda:0'), in_proj_covar=tensor([0.0321, 0.0235, 0.0206, 0.0288, 0.0227, 0.0381, 0.0298, 0.0273], + device='cuda:0'), 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:0') +2023-02-06 02:20:02,613 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6292, 2.3869, 4.3296, 1.2245, 3.0935, 2.0358, 1.9847, 2.3089], + device='cuda:0'), covar=tensor([0.1821, 0.2031, 0.0721, 0.3637, 0.1571, 0.2723, 0.1576, 0.2791], + device='cuda:0'), in_proj_covar=tensor([0.0469, 0.0460, 0.0526, 0.0535, 0.0589, 0.0523, 0.0444, 0.0590], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003], + device='cuda:0') +2023-02-06 02:20:22,673 INFO [train.py:901] (0/4) Epoch 6, batch 2800, loss[loss=0.2837, simple_loss=0.328, pruned_loss=0.1197, over 7533.00 frames. ], tot_loss[loss=0.276, simple_loss=0.3405, pruned_loss=0.1057, over 1618529.87 frames. ], batch size: 18, lr: 1.28e-02, grad_scale: 8.0 +2023-02-06 02:20:26,177 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=43221.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:20:32,056 INFO [optim.py:369] (0/4) 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 02:20:39,129 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=43239.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:20:40,378 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=43241.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:20:55,847 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 6, batch 2850, loss[loss=0.2284, simple_loss=0.3193, pruned_loss=0.06875, over 8337.00 frames. ], tot_loss[loss=0.2752, simple_loss=0.3401, pruned_loss=0.1052, over 1621510.32 frames. ], batch size: 25, lr: 1.28e-02, grad_scale: 8.0 +2023-02-06 02:21:11,140 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6425, 1.8658, 1.4584, 2.3485, 1.0911, 1.2292, 1.5310, 1.9931], + device='cuda:0'), covar=tensor([0.1058, 0.1228, 0.1549, 0.0607, 0.1494, 0.2193, 0.1356, 0.0920], + device='cuda:0'), in_proj_covar=tensor([0.0268, 0.0254, 0.0282, 0.0229, 0.0247, 0.0284, 0.0286, 0.0254], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 02:21:31,687 INFO [train.py:901] (0/4) Epoch 6, batch 2900, loss[loss=0.2787, simple_loss=0.3484, pruned_loss=0.1046, over 8357.00 frames. ], tot_loss[loss=0.2753, simple_loss=0.3403, pruned_loss=0.1052, over 1616712.09 frames. ], batch size: 24, lr: 1.28e-02, grad_scale: 8.0 +2023-02-06 02:21:41,574 INFO [optim.py:369] (0/4) 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 02:21:46,036 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43336.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:22:00,458 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43356.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:22:01,688 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 6, batch 2950, loss[loss=0.2696, simple_loss=0.3362, pruned_loss=0.1015, over 8032.00 frames. ], tot_loss[loss=0.2759, simple_loss=0.3407, pruned_loss=0.1056, over 1615371.12 frames. ], batch size: 22, lr: 1.28e-02, grad_scale: 8.0 +2023-02-06 02:22:17,903 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=43382.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:22:35,337 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43407.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:22:41,770 INFO [train.py:901] (0/4) Epoch 6, batch 3000, loss[loss=0.2497, simple_loss=0.3171, pruned_loss=0.09119, over 8074.00 frames. ], tot_loss[loss=0.2763, simple_loss=0.3407, pruned_loss=0.106, over 1609830.71 frames. ], batch size: 21, lr: 1.28e-02, grad_scale: 8.0 +2023-02-06 02:22:41,771 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 02:22:53,883 INFO [train.py:935] (0/4) Epoch 6, validation: loss=0.2158, simple_loss=0.3124, pruned_loss=0.05962, over 944034.00 frames. +2023-02-06 02:22:53,884 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6689MB +2023-02-06 02:23:03,877 INFO [optim.py:369] (0/4) 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 02:23:28,759 INFO [train.py:901] (0/4) Epoch 6, batch 3050, loss[loss=0.3791, simple_loss=0.4012, pruned_loss=0.1784, over 7107.00 frames. ], tot_loss[loss=0.2766, simple_loss=0.3414, pruned_loss=0.1059, over 1614774.46 frames. ], batch size: 71, lr: 1.28e-02, grad_scale: 8.0 +2023-02-06 02:24:03,332 INFO [train.py:901] (0/4) Epoch 6, batch 3100, loss[loss=0.2849, simple_loss=0.3492, pruned_loss=0.1103, over 7801.00 frames. ], tot_loss[loss=0.2758, simple_loss=0.3405, pruned_loss=0.1055, over 1611799.90 frames. ], batch size: 19, lr: 1.28e-02, grad_scale: 8.0 +2023-02-06 02:24:12,750 INFO [optim.py:369] (0/4) 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 02:24:14,172 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43532.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:24:21,729 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4382, 1.8674, 2.0053, 1.0802, 2.1610, 1.2127, 0.5502, 1.7197], + device='cuda:0'), covar=tensor([0.0276, 0.0129, 0.0095, 0.0231, 0.0132, 0.0429, 0.0369, 0.0119], + device='cuda:0'), in_proj_covar=tensor([0.0327, 0.0242, 0.0208, 0.0297, 0.0237, 0.0390, 0.0303, 0.0277], + device='cuda:0'), out_proj_covar=tensor([1.1037e-04, 7.9632e-05, 6.8545e-05, 9.8509e-05, 7.9991e-05, 1.4146e-04, + 1.0310e-04, 9.2497e-05], device='cuda:0') +2023-02-06 02:24:38,265 INFO [train.py:901] (0/4) Epoch 6, batch 3150, loss[loss=0.3142, simple_loss=0.3739, pruned_loss=0.1273, over 7465.00 frames. ], tot_loss[loss=0.2744, simple_loss=0.3393, pruned_loss=0.1048, over 1607313.62 frames. ], batch size: 75, lr: 1.28e-02, grad_scale: 8.0 +2023-02-06 02:24:57,055 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=43592.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:25:10,974 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=43612.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:25:13,470 INFO [train.py:901] (0/4) Epoch 6, batch 3200, loss[loss=0.2916, simple_loss=0.3481, pruned_loss=0.1175, over 8528.00 frames. ], tot_loss[loss=0.277, simple_loss=0.3416, pruned_loss=0.1062, over 1610452.38 frames. ], batch size: 49, lr: 1.28e-02, grad_scale: 8.0 +2023-02-06 02:25:14,372 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43617.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:25:23,586 INFO [optim.py:369] (0/4) 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 02:25:28,577 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43637.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:25:38,332 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 +2023-02-06 02:25:49,155 INFO [train.py:901] (0/4) Epoch 6, batch 3250, loss[loss=0.336, simple_loss=0.3667, pruned_loss=0.1527, over 8096.00 frames. ], tot_loss[loss=0.2758, simple_loss=0.3406, pruned_loss=0.1055, over 1611179.78 frames. ], batch size: 21, lr: 1.28e-02, grad_scale: 8.0 +2023-02-06 02:25:49,999 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.1771, 1.1575, 2.0276, 1.0739, 1.8823, 2.2433, 2.2241, 1.9142], + device='cuda:0'), covar=tensor([0.0889, 0.1111, 0.0537, 0.1631, 0.0657, 0.0358, 0.0494, 0.0686], + device='cuda:0'), in_proj_covar=tensor([0.0240, 0.0272, 0.0225, 0.0264, 0.0239, 0.0212, 0.0250, 0.0283], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 02:26:23,412 INFO [train.py:901] (0/4) Epoch 6, batch 3300, loss[loss=0.2921, simple_loss=0.3505, pruned_loss=0.1168, over 6596.00 frames. ], tot_loss[loss=0.2755, simple_loss=0.3404, pruned_loss=0.1053, over 1613151.50 frames. ], batch size: 71, lr: 1.28e-02, grad_scale: 8.0 +2023-02-06 02:26:33,006 INFO [optim.py:369] (0/4) 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 02:26:38,183 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-06 02:26:58,037 INFO [train.py:901] (0/4) Epoch 6, batch 3350, loss[loss=0.2681, simple_loss=0.345, pruned_loss=0.09556, over 8452.00 frames. ], tot_loss[loss=0.2747, simple_loss=0.3395, pruned_loss=0.1049, over 1612868.75 frames. ], batch size: 27, lr: 1.28e-02, grad_scale: 8.0 +2023-02-06 02:27:06,833 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3096, 2.3845, 1.6227, 1.9871, 1.9550, 1.2955, 1.7173, 1.9318], + device='cuda:0'), covar=tensor([0.1121, 0.0309, 0.0890, 0.0509, 0.0680, 0.1241, 0.0782, 0.0786], + device='cuda:0'), in_proj_covar=tensor([0.0341, 0.0232, 0.0311, 0.0302, 0.0317, 0.0312, 0.0335, 0.0309], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 02:27:25,482 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43805.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:27:31,888 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 +2023-02-06 02:27:33,330 INFO [train.py:901] (0/4) Epoch 6, batch 3400, loss[loss=0.3146, simple_loss=0.3661, pruned_loss=0.1315, over 8236.00 frames. ], tot_loss[loss=0.2763, simple_loss=0.3406, pruned_loss=0.106, over 1615643.29 frames. ], batch size: 22, lr: 1.27e-02, grad_scale: 8.0 +2023-02-06 02:27:42,440 INFO [optim.py:369] (0/4) 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 02:28:07,549 INFO [train.py:901] (0/4) Epoch 6, batch 3450, loss[loss=0.2935, simple_loss=0.3608, pruned_loss=0.1131, over 8467.00 frames. ], tot_loss[loss=0.2762, simple_loss=0.3404, pruned_loss=0.106, over 1614912.31 frames. ], batch size: 25, lr: 1.27e-02, grad_scale: 8.0 +2023-02-06 02:28:08,398 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9508, 1.4791, 3.3841, 1.4357, 2.4042, 3.7611, 3.8083, 3.1470], + device='cuda:0'), covar=tensor([0.1075, 0.1563, 0.0357, 0.1968, 0.0748, 0.0257, 0.0329, 0.0601], + device='cuda:0'), in_proj_covar=tensor([0.0241, 0.0272, 0.0224, 0.0266, 0.0237, 0.0213, 0.0249, 0.0283], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 02:28:09,127 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2027, 2.1791, 1.5656, 1.9159, 1.7984, 1.2436, 1.5871, 1.7593], + device='cuda:0'), covar=tensor([0.1046, 0.0294, 0.0827, 0.0450, 0.0570, 0.1065, 0.0858, 0.0678], + device='cuda:0'), in_proj_covar=tensor([0.0342, 0.0233, 0.0307, 0.0302, 0.0311, 0.0309, 0.0333, 0.0304], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 02:28:14,344 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=43876.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:28:17,160 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9438, 2.1962, 2.2850, 1.7176, 1.1166, 2.3415, 0.4302, 1.2748], + device='cuda:0'), covar=tensor([0.2904, 0.1779, 0.1167, 0.2263, 0.5401, 0.0945, 0.4638, 0.2373], + device='cuda:0'), in_proj_covar=tensor([0.0134, 0.0131, 0.0086, 0.0178, 0.0218, 0.0081, 0.0137, 0.0133], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 02:28:42,237 INFO [train.py:901] (0/4) Epoch 6, batch 3500, loss[loss=0.2999, simple_loss=0.3643, pruned_loss=0.1177, over 8453.00 frames. ], tot_loss[loss=0.2751, simple_loss=0.3398, pruned_loss=0.1052, over 1617154.22 frames. ], batch size: 27, lr: 1.27e-02, grad_scale: 8.0 +2023-02-06 02:28:50,513 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43927.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:28:52,405 INFO [optim.py:369] (0/4) 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 02:28:56,821 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-06 02:28:59,182 WARNING [train.py:1067] (0/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-06 02:29:16,407 INFO [train.py:901] (0/4) Epoch 6, batch 3550, loss[loss=0.2285, simple_loss=0.2955, pruned_loss=0.08077, over 7699.00 frames. ], tot_loss[loss=0.2751, simple_loss=0.34, pruned_loss=0.1051, over 1616426.91 frames. ], batch size: 18, lr: 1.27e-02, grad_scale: 8.0 +2023-02-06 02:29:24,728 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43977.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:29:34,242 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43991.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:29:40,255 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-44000.pt +2023-02-06 02:29:52,632 INFO [train.py:901] (0/4) Epoch 6, batch 3600, loss[loss=0.3068, simple_loss=0.3589, pruned_loss=0.1274, over 8293.00 frames. ], tot_loss[loss=0.2764, simple_loss=0.3408, pruned_loss=0.106, over 1615581.02 frames. ], batch size: 23, lr: 1.27e-02, grad_scale: 8.0 +2023-02-06 02:30:02,263 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 6, batch 3650, loss[loss=0.3142, simple_loss=0.3732, pruned_loss=0.1276, over 8465.00 frames. ], tot_loss[loss=0.2754, simple_loss=0.3399, pruned_loss=0.1054, over 1612008.79 frames. ], batch size: 25, lr: 1.27e-02, grad_scale: 8.0 +2023-02-06 02:31:00,588 INFO [train.py:901] (0/4) Epoch 6, batch 3700, loss[loss=0.1887, simple_loss=0.2718, pruned_loss=0.05279, over 7438.00 frames. ], tot_loss[loss=0.2739, simple_loss=0.339, pruned_loss=0.1044, over 1606344.73 frames. ], batch size: 17, lr: 1.27e-02, grad_scale: 8.0 +2023-02-06 02:31:01,281 WARNING [train.py:1067] (0/4) 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] (0/4) 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 02:31:23,982 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 6, batch 3750, loss[loss=0.2475, simple_loss=0.3053, pruned_loss=0.09483, over 7722.00 frames. ], tot_loss[loss=0.2749, simple_loss=0.3395, pruned_loss=0.1052, over 1608384.73 frames. ], batch size: 18, lr: 1.27e-02, grad_scale: 8.0 +2023-02-06 02:32:09,333 INFO [train.py:901] (0/4) Epoch 6, batch 3800, loss[loss=0.3018, simple_loss=0.3323, pruned_loss=0.1357, over 7786.00 frames. ], tot_loss[loss=0.2754, simple_loss=0.3397, pruned_loss=0.1056, over 1605988.65 frames. ], batch size: 19, lr: 1.27e-02, grad_scale: 8.0 +2023-02-06 02:32:19,582 INFO [optim.py:369] (0/4) 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 02:32:32,410 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=44247.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:32:37,978 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.6738, 4.6327, 4.1217, 1.8709, 4.0986, 4.1082, 4.2738, 3.8730], + device='cuda:0'), covar=tensor([0.0688, 0.0595, 0.0966, 0.4783, 0.0731, 0.0941, 0.1407, 0.0764], + device='cuda:0'), in_proj_covar=tensor([0.0398, 0.0298, 0.0329, 0.0411, 0.0316, 0.0289, 0.0317, 0.0262], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 02:32:44,230 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44264.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:32:45,453 INFO [train.py:901] (0/4) Epoch 6, batch 3850, loss[loss=0.3492, simple_loss=0.4012, pruned_loss=0.1485, over 8557.00 frames. ], tot_loss[loss=0.2767, simple_loss=0.3409, pruned_loss=0.1063, over 1607078.66 frames. ], batch size: 49, lr: 1.27e-02, grad_scale: 8.0 +2023-02-06 02:32:48,938 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=44271.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:32:49,759 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=44272.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:33:02,818 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-06 02:33:20,469 INFO [train.py:901] (0/4) Epoch 6, batch 3900, loss[loss=0.2656, simple_loss=0.3357, pruned_loss=0.09772, over 7935.00 frames. ], tot_loss[loss=0.277, simple_loss=0.3413, pruned_loss=0.1063, over 1610011.87 frames. ], batch size: 20, lr: 1.27e-02, grad_scale: 8.0 +2023-02-06 02:33:23,928 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=44321.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:33:30,570 INFO [optim.py:369] (0/4) 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 02:33:30,740 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0982, 2.3986, 1.7132, 2.8827, 1.3871, 1.3937, 2.0385, 2.4279], + device='cuda:0'), covar=tensor([0.0987, 0.0892, 0.1532, 0.0445, 0.1459, 0.2049, 0.1251, 0.0873], + device='cuda:0'), in_proj_covar=tensor([0.0273, 0.0251, 0.0285, 0.0225, 0.0244, 0.0279, 0.0289, 0.0258], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 02:33:56,223 INFO [train.py:901] (0/4) Epoch 6, batch 3950, loss[loss=0.2314, simple_loss=0.2953, pruned_loss=0.08375, over 7701.00 frames. ], tot_loss[loss=0.2763, simple_loss=0.3408, pruned_loss=0.1059, over 1612411.39 frames. ], batch size: 18, lr: 1.27e-02, grad_scale: 8.0 +2023-02-06 02:34:09,756 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44386.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:34:17,206 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3087, 1.9495, 3.1678, 2.4278, 2.6979, 2.0215, 1.5555, 1.3869], + device='cuda:0'), covar=tensor([0.2222, 0.2469, 0.0555, 0.1447, 0.1147, 0.1324, 0.1243, 0.2666], + device='cuda:0'), in_proj_covar=tensor([0.0796, 0.0729, 0.0626, 0.0719, 0.0813, 0.0668, 0.0636, 0.0666], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 02:34:30,790 INFO [train.py:901] (0/4) Epoch 6, batch 4000, loss[loss=0.2361, simple_loss=0.3184, pruned_loss=0.07689, over 8247.00 frames. ], tot_loss[loss=0.2763, simple_loss=0.3411, pruned_loss=0.1057, over 1615017.92 frames. ], batch size: 24, lr: 1.27e-02, grad_scale: 8.0 +2023-02-06 02:34:40,323 INFO [optim.py:369] (0/4) 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 02:34:43,276 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4453, 1.2444, 1.3503, 1.1638, 0.8556, 1.1879, 1.1671, 1.0359], + device='cuda:0'), covar=tensor([0.0588, 0.1332, 0.1781, 0.1454, 0.0610, 0.1601, 0.0753, 0.0650], + device='cuda:0'), in_proj_covar=tensor([0.0125, 0.0173, 0.0214, 0.0178, 0.0122, 0.0182, 0.0137, 0.0149], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0007, 0.0008, 0.0007, 0.0005, 0.0007, 0.0006, 0.0006], + device='cuda:0') +2023-02-06 02:34:44,595 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44436.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:35:05,725 INFO [train.py:901] (0/4) Epoch 6, batch 4050, loss[loss=0.2842, simple_loss=0.3623, pruned_loss=0.103, over 8539.00 frames. ], tot_loss[loss=0.2759, simple_loss=0.3406, pruned_loss=0.1056, over 1614442.80 frames. ], batch size: 28, lr: 1.27e-02, grad_scale: 16.0 +2023-02-06 02:35:06,074 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 +2023-02-06 02:35:06,869 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.52 vs. limit=5.0 +2023-02-06 02:35:41,190 INFO [train.py:901] (0/4) Epoch 6, batch 4100, loss[loss=0.2293, simple_loss=0.3002, pruned_loss=0.07923, over 8084.00 frames. ], tot_loss[loss=0.2753, simple_loss=0.3398, pruned_loss=0.1054, over 1612651.71 frames. ], batch size: 21, lr: 1.26e-02, grad_scale: 16.0 +2023-02-06 02:35:44,032 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=44520.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:35:50,486 INFO [optim.py:369] (0/4) 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 02:36:00,537 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44545.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:36:00,598 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=44545.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:36:14,412 INFO [train.py:901] (0/4) Epoch 6, batch 4150, loss[loss=0.2621, simple_loss=0.3333, pruned_loss=0.09546, over 8297.00 frames. ], tot_loss[loss=0.2747, simple_loss=0.3393, pruned_loss=0.105, over 1612012.16 frames. ], batch size: 23, lr: 1.26e-02, grad_scale: 16.0 +2023-02-06 02:36:15,774 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44568.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:36:49,812 INFO [train.py:901] (0/4) Epoch 6, batch 4200, loss[loss=0.2643, simple_loss=0.3313, pruned_loss=0.09864, over 8247.00 frames. ], tot_loss[loss=0.2727, simple_loss=0.3377, pruned_loss=0.1039, over 1609759.40 frames. ], batch size: 22, lr: 1.26e-02, grad_scale: 16.0 +2023-02-06 02:36:58,994 INFO [optim.py:369] (0/4) 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 02:37:05,653 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-06 02:37:07,964 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=44642.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:37:23,784 INFO [train.py:901] (0/4) Epoch 6, batch 4250, loss[loss=0.3014, simple_loss=0.3612, pruned_loss=0.1208, over 8365.00 frames. ], tot_loss[loss=0.2715, simple_loss=0.3371, pruned_loss=0.103, over 1607110.89 frames. ], batch size: 24, lr: 1.26e-02, grad_scale: 16.0 +2023-02-06 02:37:24,678 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=44667.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:37:29,276 WARNING [train.py:1067] (0/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-06 02:37:41,577 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=44692.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:37:58,483 INFO [train.py:901] (0/4) Epoch 6, batch 4300, loss[loss=0.2505, simple_loss=0.3202, pruned_loss=0.09039, over 7921.00 frames. ], tot_loss[loss=0.2709, simple_loss=0.3367, pruned_loss=0.1025, over 1606185.83 frames. ], batch size: 20, lr: 1.26e-02, grad_scale: 16.0 +2023-02-06 02:38:00,029 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=44717.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:38:05,900 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 +2023-02-06 02:38:08,670 INFO [optim.py:369] (0/4) 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 02:38:28,323 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 +2023-02-06 02:38:31,335 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8500, 3.8957, 2.4172, 2.6669, 3.2088, 1.9976, 2.5030, 3.1630], + device='cuda:0'), covar=tensor([0.1432, 0.0269, 0.0882, 0.0743, 0.0620, 0.1179, 0.0934, 0.0865], + device='cuda:0'), in_proj_covar=tensor([0.0345, 0.0233, 0.0308, 0.0304, 0.0316, 0.0312, 0.0331, 0.0309], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 02:38:33,200 INFO [train.py:901] (0/4) Epoch 6, batch 4350, loss[loss=0.2643, simple_loss=0.3323, pruned_loss=0.0982, over 8078.00 frames. ], tot_loss[loss=0.2728, simple_loss=0.3375, pruned_loss=0.1041, over 1605811.37 frames. ], batch size: 21, lr: 1.26e-02, grad_scale: 16.0 +2023-02-06 02:39:00,034 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 6, batch 4400, loss[loss=0.2568, simple_loss=0.3163, pruned_loss=0.09866, over 6795.00 frames. ], tot_loss[loss=0.2739, simple_loss=0.3382, pruned_loss=0.1048, over 1604701.43 frames. ], batch size: 15, lr: 1.26e-02, grad_scale: 16.0 +2023-02-06 02:39:16,200 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1749, 1.8086, 2.9245, 2.3469, 2.4467, 1.8040, 1.3612, 1.1289], + device='cuda:0'), covar=tensor([0.2333, 0.2553, 0.0581, 0.1268, 0.1001, 0.1440, 0.1474, 0.2429], + device='cuda:0'), in_proj_covar=tensor([0.0783, 0.0724, 0.0619, 0.0716, 0.0806, 0.0658, 0.0629, 0.0657], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 02:39:17,273 INFO [optim.py:369] (0/4) 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 02:39:40,219 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 6, batch 4450, loss[loss=0.2512, simple_loss=0.3098, pruned_loss=0.09634, over 7534.00 frames. ], tot_loss[loss=0.2741, simple_loss=0.3379, pruned_loss=0.1052, over 1600141.13 frames. ], batch size: 18, lr: 1.26e-02, grad_scale: 16.0 +2023-02-06 02:39:58,549 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=44889.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:40:13,885 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 6, batch 4500, loss[loss=0.3435, simple_loss=0.3793, pruned_loss=0.1539, over 6619.00 frames. ], tot_loss[loss=0.2754, simple_loss=0.339, pruned_loss=0.1058, over 1602132.24 frames. ], batch size: 71, lr: 1.26e-02, grad_scale: 16.0 +2023-02-06 02:40:26,436 INFO [optim.py:369] (0/4) 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 02:40:31,783 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-06 02:40:44,433 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 +2023-02-06 02:40:52,034 INFO [train.py:901] (0/4) Epoch 6, batch 4550, loss[loss=0.3362, simple_loss=0.3842, pruned_loss=0.1441, over 8576.00 frames. ], tot_loss[loss=0.277, simple_loss=0.3404, pruned_loss=0.1068, over 1602687.53 frames. ], batch size: 39, lr: 1.26e-02, grad_scale: 16.0 +2023-02-06 02:41:16,644 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.2744, 1.3253, 4.5471, 1.7268, 3.8007, 3.7285, 4.0324, 3.8828], + device='cuda:0'), covar=tensor([0.0488, 0.3633, 0.0355, 0.2419, 0.1159, 0.0737, 0.0472, 0.0571], + device='cuda:0'), in_proj_covar=tensor([0.0360, 0.0501, 0.0457, 0.0440, 0.0512, 0.0421, 0.0423, 0.0480], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 02:41:18,792 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45004.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:41:27,057 INFO [train.py:901] (0/4) Epoch 6, batch 4600, loss[loss=0.1941, simple_loss=0.2703, pruned_loss=0.05899, over 7438.00 frames. ], tot_loss[loss=0.2777, simple_loss=0.3411, pruned_loss=0.1071, over 1605287.37 frames. ], batch size: 17, lr: 1.26e-02, grad_scale: 16.0 +2023-02-06 02:41:34,812 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45027.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:41:36,654 INFO [optim.py:369] (0/4) 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 02:42:02,750 INFO [train.py:901] (0/4) Epoch 6, batch 4650, loss[loss=0.3227, simple_loss=0.3815, pruned_loss=0.1319, over 8125.00 frames. ], tot_loss[loss=0.2774, simple_loss=0.3412, pruned_loss=0.1068, over 1606369.91 frames. ], batch size: 22, lr: 1.26e-02, grad_scale: 16.0 +2023-02-06 02:42:08,367 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=45074.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:42:25,260 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=45099.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:42:36,888 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-02-06 02:42:36,979 INFO [train.py:901] (0/4) Epoch 6, batch 4700, loss[loss=0.2263, simple_loss=0.3069, pruned_loss=0.07279, over 8091.00 frames. ], tot_loss[loss=0.2755, simple_loss=0.3399, pruned_loss=0.1055, over 1609163.82 frames. ], batch size: 21, lr: 1.26e-02, grad_scale: 16.0 +2023-02-06 02:42:46,395 INFO [optim.py:369] (0/4) 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 02:42:49,953 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.87 vs. limit=2.0 +2023-02-06 02:43:11,110 INFO [train.py:901] (0/4) Epoch 6, batch 4750, loss[loss=0.2993, simple_loss=0.3541, pruned_loss=0.1222, over 8148.00 frames. ], tot_loss[loss=0.2768, simple_loss=0.3409, pruned_loss=0.1064, over 1610131.85 frames. ], batch size: 22, lr: 1.26e-02, grad_scale: 16.0 +2023-02-06 02:43:30,409 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-06 02:43:31,798 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-06 02:43:45,096 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6437, 1.9420, 2.3025, 1.2881, 2.4718, 1.4752, 0.7486, 1.8958], + device='cuda:0'), covar=tensor([0.0297, 0.0140, 0.0104, 0.0212, 0.0120, 0.0415, 0.0338, 0.0111], + device='cuda:0'), in_proj_covar=tensor([0.0330, 0.0237, 0.0205, 0.0297, 0.0236, 0.0381, 0.0302, 0.0280], + device='cuda:0'), out_proj_covar=tensor([1.1020e-04, 7.6743e-05, 6.6699e-05, 9.7493e-05, 7.8530e-05, 1.3569e-04, + 1.0132e-04, 9.2449e-05], device='cuda:0') +2023-02-06 02:43:46,229 INFO [train.py:901] (0/4) Epoch 6, batch 4800, loss[loss=0.1996, simple_loss=0.2659, pruned_loss=0.06663, over 7691.00 frames. ], tot_loss[loss=0.2782, simple_loss=0.342, pruned_loss=0.1072, over 1612904.21 frames. ], batch size: 18, lr: 1.26e-02, grad_scale: 16.0 +2023-02-06 02:43:50,544 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.1774, 1.1960, 1.1921, 1.1342, 0.8506, 1.2346, 0.1009, 1.0448], + device='cuda:0'), covar=tensor([0.3323, 0.2161, 0.1009, 0.2118, 0.5674, 0.0902, 0.4811, 0.2017], + device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0136, 0.0088, 0.0184, 0.0226, 0.0084, 0.0146, 0.0138], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 02:43:55,771 INFO [optim.py:369] (0/4) 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 02:44:16,239 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 6, batch 4850, loss[loss=0.2341, simple_loss=0.2843, pruned_loss=0.09195, over 5957.00 frames. ], tot_loss[loss=0.2778, simple_loss=0.3418, pruned_loss=0.1069, over 1614193.04 frames. ], batch size: 13, lr: 1.25e-02, grad_scale: 16.0 +2023-02-06 02:44:20,857 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-06 02:44:22,378 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3746, 1.7338, 1.8680, 1.0743, 2.0438, 1.2740, 0.3945, 1.5940], + device='cuda:0'), covar=tensor([0.0249, 0.0141, 0.0111, 0.0187, 0.0113, 0.0422, 0.0364, 0.0115], + device='cuda:0'), in_proj_covar=tensor([0.0329, 0.0239, 0.0206, 0.0297, 0.0236, 0.0382, 0.0304, 0.0283], + device='cuda:0'), out_proj_covar=tensor([1.0989e-04, 7.7520e-05, 6.6989e-05, 9.7550e-05, 7.8243e-05, 1.3595e-04, + 1.0183e-04, 9.3360e-05], device='cuda:0') +2023-02-06 02:44:28,390 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.2630, 1.7286, 1.6569, 1.8033, 1.7918, 1.7934, 2.4504, 1.7094], + device='cuda:0'), covar=tensor([0.0460, 0.1297, 0.1784, 0.1271, 0.0534, 0.1542, 0.0658, 0.0612], + device='cuda:0'), in_proj_covar=tensor([0.0126, 0.0175, 0.0216, 0.0178, 0.0123, 0.0183, 0.0138, 0.0151], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0007, 0.0008, 0.0008, 0.0005, 0.0008, 0.0006, 0.0006], + device='cuda:0') +2023-02-06 02:44:32,436 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45283.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:44:35,152 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45285.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:44:50,592 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45308.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:44:56,003 INFO [train.py:901] (0/4) Epoch 6, batch 4900, loss[loss=0.2694, simple_loss=0.3352, pruned_loss=0.1018, over 8232.00 frames. ], tot_loss[loss=0.2772, simple_loss=0.3413, pruned_loss=0.1066, over 1606783.56 frames. ], batch size: 22, lr: 1.25e-02, grad_scale: 16.0 +2023-02-06 02:45:05,436 INFO [optim.py:369] (0/4) 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 02:45:13,764 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.99 vs. limit=5.0 +2023-02-06 02:45:30,238 INFO [train.py:901] (0/4) Epoch 6, batch 4950, loss[loss=0.2802, simple_loss=0.3375, pruned_loss=0.1114, over 7924.00 frames. ], tot_loss[loss=0.2769, simple_loss=0.3413, pruned_loss=0.1062, over 1609655.29 frames. ], batch size: 20, lr: 1.25e-02, grad_scale: 16.0 +2023-02-06 02:45:30,381 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=45366.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 02:46:05,717 INFO [train.py:901] (0/4) Epoch 6, batch 5000, loss[loss=0.2889, simple_loss=0.3597, pruned_loss=0.1091, over 8354.00 frames. ], tot_loss[loss=0.2774, simple_loss=0.3416, pruned_loss=0.1066, over 1614451.08 frames. ], batch size: 24, lr: 1.25e-02, grad_scale: 16.0 +2023-02-06 02:46:07,204 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=45418.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:46:08,576 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3390, 2.2648, 1.6723, 2.1644, 1.7280, 1.3887, 1.6624, 1.7869], + device='cuda:0'), covar=tensor([0.1048, 0.0284, 0.0873, 0.0381, 0.0723, 0.1202, 0.0715, 0.0608], + device='cuda:0'), in_proj_covar=tensor([0.0341, 0.0228, 0.0305, 0.0297, 0.0314, 0.0311, 0.0329, 0.0301], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 02:46:15,101 INFO [optim.py:369] (0/4) 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 02:46:24,042 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=45443.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:46:40,013 INFO [train.py:901] (0/4) Epoch 6, batch 5050, loss[loss=0.2134, simple_loss=0.3045, pruned_loss=0.06114, over 8446.00 frames. ], tot_loss[loss=0.2752, simple_loss=0.3397, pruned_loss=0.1054, over 1612926.12 frames. ], batch size: 25, lr: 1.25e-02, grad_scale: 16.0 +2023-02-06 02:46:49,880 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 +2023-02-06 02:46:58,882 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-06 02:47:14,058 INFO [train.py:901] (0/4) Epoch 6, batch 5100, loss[loss=0.2544, simple_loss=0.3317, pruned_loss=0.0886, over 8481.00 frames. ], tot_loss[loss=0.2756, simple_loss=0.3404, pruned_loss=0.1055, over 1615791.85 frames. ], batch size: 25, lr: 1.25e-02, grad_scale: 16.0 +2023-02-06 02:47:24,712 INFO [optim.py:369] (0/4) 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 02:47:26,958 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45533.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:47:43,433 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45558.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:47:49,305 INFO [train.py:901] (0/4) Epoch 6, batch 5150, loss[loss=0.317, simple_loss=0.3772, pruned_loss=0.1284, over 8327.00 frames. ], tot_loss[loss=0.275, simple_loss=0.3398, pruned_loss=0.1051, over 1615610.54 frames. ], batch size: 26, lr: 1.25e-02, grad_scale: 16.0 +2023-02-06 02:48:23,748 INFO [train.py:901] (0/4) Epoch 6, batch 5200, loss[loss=0.2689, simple_loss=0.3366, pruned_loss=0.1006, over 8595.00 frames. ], tot_loss[loss=0.2726, simple_loss=0.3377, pruned_loss=0.1038, over 1611925.13 frames. ], batch size: 31, lr: 1.25e-02, grad_scale: 16.0 +2023-02-06 02:48:34,003 INFO [optim.py:369] (0/4) 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 02:48:38,460 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7229, 1.9602, 2.2337, 1.6370, 1.1249, 2.2124, 0.5444, 1.3733], + device='cuda:0'), covar=tensor([0.2637, 0.1758, 0.0623, 0.2557, 0.5717, 0.0490, 0.4592, 0.2215], + device='cuda:0'), in_proj_covar=tensor([0.0138, 0.0137, 0.0088, 0.0184, 0.0228, 0.0083, 0.0148, 0.0138], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 02:48:57,683 WARNING [train.py:1067] (0/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-06 02:48:59,759 INFO [train.py:901] (0/4) Epoch 6, batch 5250, loss[loss=0.224, simple_loss=0.3016, pruned_loss=0.07321, over 8145.00 frames. ], tot_loss[loss=0.2736, simple_loss=0.3383, pruned_loss=0.1045, over 1611501.00 frames. ], batch size: 22, lr: 1.25e-02, grad_scale: 16.0 +2023-02-06 02:49:30,199 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=45710.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 02:49:34,102 INFO [train.py:901] (0/4) Epoch 6, batch 5300, loss[loss=0.3018, simple_loss=0.3624, pruned_loss=0.1206, over 8232.00 frames. ], tot_loss[loss=0.2754, simple_loss=0.3393, pruned_loss=0.1057, over 1607537.75 frames. ], batch size: 24, lr: 1.25e-02, grad_scale: 16.0 +2023-02-06 02:49:36,937 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=45720.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:49:43,557 INFO [optim.py:369] (0/4) 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 02:50:09,986 INFO [train.py:901] (0/4) Epoch 6, batch 5350, loss[loss=0.2919, simple_loss=0.36, pruned_loss=0.1119, over 8030.00 frames. ], tot_loss[loss=0.2764, simple_loss=0.3408, pruned_loss=0.106, over 1615833.63 frames. ], batch size: 22, lr: 1.25e-02, grad_scale: 16.0 +2023-02-06 02:50:25,436 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45789.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:50:27,326 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=45792.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:50:42,756 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45814.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:50:42,775 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45814.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:50:43,910 INFO [train.py:901] (0/4) Epoch 6, batch 5400, loss[loss=0.307, simple_loss=0.3586, pruned_loss=0.1277, over 8343.00 frames. ], tot_loss[loss=0.2768, simple_loss=0.3415, pruned_loss=0.1061, over 1619171.06 frames. ], batch size: 26, lr: 1.25e-02, grad_scale: 16.0 +2023-02-06 02:50:49,971 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45825.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 02:50:53,700 INFO [optim.py:369] (0/4) 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 02:50:59,021 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45839.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:51:00,972 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0159, 1.6646, 3.3382, 1.5024, 2.3198, 3.7987, 3.7311, 3.3311], + device='cuda:0'), covar=tensor([0.0991, 0.1293, 0.0399, 0.1781, 0.0757, 0.0254, 0.0385, 0.0535], + device='cuda:0'), in_proj_covar=tensor([0.0239, 0.0266, 0.0229, 0.0267, 0.0236, 0.0216, 0.0251, 0.0276], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 02:51:17,302 INFO [train.py:901] (0/4) Epoch 6, batch 5450, loss[loss=0.2383, simple_loss=0.2971, pruned_loss=0.08977, over 7240.00 frames. ], tot_loss[loss=0.2747, simple_loss=0.3399, pruned_loss=0.1048, over 1616925.24 frames. ], batch size: 16, lr: 1.25e-02, grad_scale: 8.0 +2023-02-06 02:51:25,559 INFO [zipformer.py:1185] (0/4) 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] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-06 02:51:52,367 INFO [train.py:901] (0/4) Epoch 6, batch 5500, loss[loss=0.3354, simple_loss=0.3825, pruned_loss=0.1442, over 8538.00 frames. ], tot_loss[loss=0.275, simple_loss=0.3402, pruned_loss=0.1049, over 1612231.88 frames. ], batch size: 31, lr: 1.25e-02, grad_scale: 8.0 +2023-02-06 02:52:03,083 INFO [optim.py:369] (0/4) 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 02:52:05,380 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4765, 1.8048, 3.3043, 1.1451, 2.3504, 1.8859, 1.4742, 2.0124], + device='cuda:0'), covar=tensor([0.1459, 0.1885, 0.0606, 0.3161, 0.1262, 0.2307, 0.1543, 0.1844], + device='cuda:0'), in_proj_covar=tensor([0.0466, 0.0458, 0.0524, 0.0540, 0.0583, 0.0530, 0.0443, 0.0581], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003], + device='cuda:0') +2023-02-06 02:52:27,011 INFO [train.py:901] (0/4) Epoch 6, batch 5550, loss[loss=0.2341, simple_loss=0.2988, pruned_loss=0.08469, over 7927.00 frames. ], tot_loss[loss=0.274, simple_loss=0.3389, pruned_loss=0.1045, over 1609140.33 frames. ], batch size: 20, lr: 1.25e-02, grad_scale: 8.0 +2023-02-06 02:52:51,788 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-46000.pt +2023-02-06 02:53:03,509 INFO [train.py:901] (0/4) Epoch 6, batch 5600, loss[loss=0.2411, simple_loss=0.3184, pruned_loss=0.08196, over 8026.00 frames. ], tot_loss[loss=0.2733, simple_loss=0.3387, pruned_loss=0.1039, over 1612167.82 frames. ], batch size: 22, lr: 1.24e-02, grad_scale: 8.0 +2023-02-06 02:53:04,342 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3146, 2.0938, 3.8205, 2.0382, 2.3930, 4.3186, 4.2157, 3.8478], + device='cuda:0'), covar=tensor([0.0959, 0.1188, 0.0407, 0.1683, 0.0941, 0.0259, 0.0364, 0.0509], + device='cuda:0'), in_proj_covar=tensor([0.0235, 0.0264, 0.0226, 0.0264, 0.0236, 0.0212, 0.0248, 0.0272], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:0') +2023-02-06 02:53:13,360 INFO [optim.py:369] (0/4) 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 02:53:25,613 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3040, 2.1321, 1.6670, 1.9317, 1.8111, 1.3061, 1.6315, 1.7806], + device='cuda:0'), covar=tensor([0.1153, 0.0330, 0.0935, 0.0502, 0.0572, 0.1226, 0.0796, 0.0755], + device='cuda:0'), in_proj_covar=tensor([0.0355, 0.0237, 0.0314, 0.0307, 0.0315, 0.0318, 0.0341, 0.0314], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 02:53:35,998 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=46064.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:53:37,297 INFO [train.py:901] (0/4) Epoch 6, batch 5650, loss[loss=0.2829, simple_loss=0.3454, pruned_loss=0.1102, over 8611.00 frames. ], tot_loss[loss=0.2723, simple_loss=0.3382, pruned_loss=0.1032, over 1614021.93 frames. ], batch size: 34, lr: 1.24e-02, grad_scale: 8.0 +2023-02-06 02:53:47,403 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46081.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 02:53:51,162 WARNING [train.py:1067] (0/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-06 02:54:04,722 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46106.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 02:54:12,521 INFO [train.py:901] (0/4) Epoch 6, batch 5700, loss[loss=0.2795, simple_loss=0.3418, pruned_loss=0.1086, over 8685.00 frames. ], tot_loss[loss=0.2728, simple_loss=0.3383, pruned_loss=0.1037, over 1612739.93 frames. ], batch size: 34, lr: 1.24e-02, grad_scale: 8.0 +2023-02-06 02:54:22,533 INFO [optim.py:369] (0/4) 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 02:54:25,965 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=46136.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:54:32,807 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.7818, 2.2232, 3.9637, 2.9355, 3.1738, 2.3534, 1.7966, 1.4973], + device='cuda:0'), covar=tensor([0.2191, 0.2693, 0.0561, 0.1558, 0.1249, 0.1189, 0.1117, 0.2923], + device='cuda:0'), in_proj_covar=tensor([0.0800, 0.0736, 0.0626, 0.0725, 0.0830, 0.0676, 0.0640, 0.0676], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 02:54:46,332 INFO [train.py:901] (0/4) Epoch 6, batch 5750, loss[loss=0.2719, simple_loss=0.3342, pruned_loss=0.1048, over 8281.00 frames. ], tot_loss[loss=0.2727, simple_loss=0.338, pruned_loss=0.1037, over 1613010.64 frames. ], batch size: 23, lr: 1.24e-02, grad_scale: 8.0 +2023-02-06 02:54:50,064 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 +2023-02-06 02:54:53,644 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-06 02:54:55,254 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46179.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:55:04,383 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.89 vs. limit=2.0 +2023-02-06 02:55:21,188 INFO [train.py:901] (0/4) Epoch 6, batch 5800, loss[loss=0.374, simple_loss=0.4203, pruned_loss=0.1639, over 7049.00 frames. ], tot_loss[loss=0.2713, simple_loss=0.3372, pruned_loss=0.1027, over 1613782.50 frames. ], batch size: 72, lr: 1.24e-02, grad_scale: 8.0 +2023-02-06 02:55:24,805 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=46221.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:55:32,618 INFO [optim.py:369] (0/4) 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 02:55:45,808 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46251.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:55:55,778 INFO [train.py:901] (0/4) Epoch 6, batch 5850, loss[loss=0.2316, simple_loss=0.2944, pruned_loss=0.08441, over 7780.00 frames. ], tot_loss[loss=0.2713, simple_loss=0.3374, pruned_loss=0.1026, over 1612204.49 frames. ], batch size: 19, lr: 1.24e-02, grad_scale: 8.0 +2023-02-06 02:56:14,970 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46294.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:56:29,466 INFO [train.py:901] (0/4) Epoch 6, batch 5900, loss[loss=0.3047, simple_loss=0.362, pruned_loss=0.1237, over 8365.00 frames. ], tot_loss[loss=0.2701, simple_loss=0.3359, pruned_loss=0.1021, over 1609434.37 frames. ], batch size: 49, lr: 1.24e-02, grad_scale: 8.0 +2023-02-06 02:56:39,473 INFO [optim.py:369] (0/4) 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 02:56:43,676 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46336.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:57:04,654 INFO [train.py:901] (0/4) Epoch 6, batch 5950, loss[loss=0.1972, simple_loss=0.2684, pruned_loss=0.063, over 7438.00 frames. ], tot_loss[loss=0.2691, simple_loss=0.3353, pruned_loss=0.1014, over 1611971.31 frames. ], batch size: 17, lr: 1.24e-02, grad_scale: 8.0 +2023-02-06 02:57:38,429 INFO [train.py:901] (0/4) Epoch 6, batch 6000, loss[loss=0.2308, simple_loss=0.3049, pruned_loss=0.07837, over 8074.00 frames. ], tot_loss[loss=0.2692, simple_loss=0.3351, pruned_loss=0.1017, over 1611500.87 frames. ], batch size: 21, lr: 1.24e-02, grad_scale: 8.0 +2023-02-06 02:57:38,430 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 02:57:50,763 INFO [train.py:935] (0/4) Epoch 6, validation: loss=0.2127, simple_loss=0.3094, pruned_loss=0.05799, over 944034.00 frames. +2023-02-06 02:57:50,764 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6689MB +2023-02-06 02:57:55,753 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.2931, 4.3170, 3.8438, 1.9515, 3.7462, 3.9235, 4.0652, 3.3457], + device='cuda:0'), covar=tensor([0.0929, 0.0610, 0.1127, 0.5007, 0.0828, 0.0744, 0.1284, 0.0887], + device='cuda:0'), in_proj_covar=tensor([0.0414, 0.0313, 0.0341, 0.0431, 0.0332, 0.0305, 0.0327, 0.0274], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 02:58:01,257 INFO [optim.py:369] (0/4) 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 02:58:04,332 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46435.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:58:21,754 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46460.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:58:25,654 INFO [train.py:901] (0/4) Epoch 6, batch 6050, loss[loss=0.2144, simple_loss=0.299, pruned_loss=0.06489, over 8234.00 frames. ], tot_loss[loss=0.268, simple_loss=0.3342, pruned_loss=0.1009, over 1607945.22 frames. ], batch size: 22, lr: 1.24e-02, grad_scale: 8.0 +2023-02-06 02:58:56,091 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46507.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:59:02,111 INFO [train.py:901] (0/4) Epoch 6, batch 6100, loss[loss=0.2924, simple_loss=0.3403, pruned_loss=0.1222, over 7534.00 frames. ], tot_loss[loss=0.2692, simple_loss=0.3351, pruned_loss=0.1017, over 1606521.38 frames. ], batch size: 18, lr: 1.24e-02, grad_scale: 8.0 +2023-02-06 02:59:12,641 INFO [optim.py:369] (0/4) 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 02:59:13,543 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46532.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 02:59:17,679 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4796, 1.4411, 1.5814, 1.3226, 1.1755, 1.4056, 1.7626, 1.5459], + device='cuda:0'), covar=tensor([0.0490, 0.1254, 0.1679, 0.1369, 0.0604, 0.1488, 0.0694, 0.0581], + device='cuda:0'), in_proj_covar=tensor([0.0122, 0.0171, 0.0211, 0.0173, 0.0121, 0.0180, 0.0134, 0.0147], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0007, 0.0008, 0.0007, 0.0005, 0.0007, 0.0006, 0.0006], + device='cuda:0') +2023-02-06 02:59:24,580 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-06 02:59:37,759 INFO [train.py:901] (0/4) Epoch 6, batch 6150, loss[loss=0.3216, simple_loss=0.3679, pruned_loss=0.1377, over 8339.00 frames. ], tot_loss[loss=0.2702, simple_loss=0.336, pruned_loss=0.1023, over 1608445.60 frames. ], batch size: 26, lr: 1.24e-02, grad_scale: 8.0 +2023-02-06 02:59:43,344 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.7742, 1.7185, 5.8813, 2.0591, 5.2638, 4.8781, 5.4610, 5.3367], + device='cuda:0'), covar=tensor([0.0492, 0.3819, 0.0311, 0.2589, 0.0945, 0.0641, 0.0421, 0.0497], + device='cuda:0'), in_proj_covar=tensor([0.0366, 0.0517, 0.0468, 0.0445, 0.0518, 0.0428, 0.0437, 0.0481], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 02:59:56,241 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46592.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:00:11,856 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46613.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:00:13,815 INFO [train.py:901] (0/4) Epoch 6, batch 6200, loss[loss=0.2753, simple_loss=0.337, pruned_loss=0.1068, over 8258.00 frames. ], tot_loss[loss=0.2728, simple_loss=0.3383, pruned_loss=0.1036, over 1612940.81 frames. ], batch size: 24, lr: 1.24e-02, grad_scale: 8.0 +2023-02-06 03:00:14,710 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46617.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:00:16,761 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46620.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:00:24,151 INFO [optim.py:369] (0/4) 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 03:00:28,935 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=46638.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:00:48,690 INFO [train.py:901] (0/4) Epoch 6, batch 6250, loss[loss=0.2089, simple_loss=0.2788, pruned_loss=0.06946, over 7524.00 frames. ], tot_loss[loss=0.2705, simple_loss=0.3361, pruned_loss=0.1024, over 1609402.09 frames. ], batch size: 18, lr: 1.24e-02, grad_scale: 8.0 +2023-02-06 03:01:08,350 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6979, 1.9951, 1.5969, 2.4730, 1.1712, 1.2381, 1.6623, 2.0660], + device='cuda:0'), covar=tensor([0.1117, 0.1131, 0.1556, 0.0683, 0.1480, 0.1987, 0.1386, 0.1001], + device='cuda:0'), in_proj_covar=tensor([0.0269, 0.0251, 0.0282, 0.0228, 0.0248, 0.0278, 0.0283, 0.0248], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 03:01:22,870 INFO [train.py:901] (0/4) Epoch 6, batch 6300, loss[loss=0.2819, simple_loss=0.3462, pruned_loss=0.1087, over 8335.00 frames. ], tot_loss[loss=0.2718, simple_loss=0.3377, pruned_loss=0.1029, over 1612958.78 frames. ], batch size: 26, lr: 1.24e-02, grad_scale: 8.0 +2023-02-06 03:01:34,428 INFO [optim.py:369] (0/4) 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 03:01:49,390 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46753.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:01:57,778 INFO [train.py:901] (0/4) Epoch 6, batch 6350, loss[loss=0.2701, simple_loss=0.3517, pruned_loss=0.09423, over 8360.00 frames. ], tot_loss[loss=0.2712, simple_loss=0.3372, pruned_loss=0.1026, over 1615053.54 frames. ], batch size: 24, lr: 1.23e-02, grad_scale: 8.0 +2023-02-06 03:02:32,254 INFO [train.py:901] (0/4) Epoch 6, batch 6400, loss[loss=0.2995, simple_loss=0.3635, pruned_loss=0.1178, over 8656.00 frames. ], tot_loss[loss=0.273, simple_loss=0.3386, pruned_loss=0.1037, over 1614250.61 frames. ], batch size: 34, lr: 1.23e-02, grad_scale: 8.0 +2023-02-06 03:02:41,187 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46828.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:02:43,100 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 6, batch 6450, loss[loss=0.325, simple_loss=0.3678, pruned_loss=0.1411, over 8478.00 frames. ], tot_loss[loss=0.2738, simple_loss=0.3396, pruned_loss=0.1039, over 1618391.63 frames. ], batch size: 25, lr: 1.23e-02, grad_scale: 8.0 +2023-02-06 03:03:09,170 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.30 vs. limit=5.0 +2023-02-06 03:03:41,568 INFO [train.py:901] (0/4) Epoch 6, batch 6500, loss[loss=0.2561, simple_loss=0.333, pruned_loss=0.08965, over 8100.00 frames. ], tot_loss[loss=0.2745, simple_loss=0.3397, pruned_loss=0.1047, over 1618794.99 frames. ], batch size: 23, lr: 1.23e-02, grad_scale: 8.0 +2023-02-06 03:03:51,600 INFO [optim.py:369] (0/4) 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 03:04:09,436 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=46957.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:04:14,546 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 6, batch 6550, loss[loss=0.3429, simple_loss=0.3776, pruned_loss=0.1541, over 7172.00 frames. ], tot_loss[loss=0.2762, simple_loss=0.3413, pruned_loss=0.1056, over 1617123.89 frames. ], batch size: 72, lr: 1.23e-02, grad_scale: 8.0 +2023-02-06 03:04:37,514 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-06 03:04:45,791 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47009.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:04:50,972 INFO [train.py:901] (0/4) Epoch 6, batch 6600, loss[loss=0.2392, simple_loss=0.3243, pruned_loss=0.07703, over 8107.00 frames. ], tot_loss[loss=0.2749, simple_loss=0.3403, pruned_loss=0.1048, over 1614722.06 frames. ], batch size: 23, lr: 1.23e-02, grad_scale: 8.0 +2023-02-06 03:04:56,464 WARNING [train.py:1067] (0/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-06 03:05:01,161 INFO [optim.py:369] (0/4) 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 03:05:03,290 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47034.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:05:14,290 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 +2023-02-06 03:05:15,514 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.63 vs. limit=2.0 +2023-02-06 03:05:25,532 INFO [train.py:901] (0/4) Epoch 6, batch 6650, loss[loss=0.2881, simple_loss=0.3469, pruned_loss=0.1146, over 7967.00 frames. ], tot_loss[loss=0.2757, simple_loss=0.3407, pruned_loss=0.1054, over 1613463.26 frames. ], batch size: 21, lr: 1.23e-02, grad_scale: 4.0 +2023-02-06 03:05:30,595 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47072.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:05:35,410 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47079.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:05:45,855 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 +2023-02-06 03:06:00,869 INFO [train.py:901] (0/4) Epoch 6, batch 6700, loss[loss=0.2561, simple_loss=0.3247, pruned_loss=0.0937, over 8098.00 frames. ], tot_loss[loss=0.2756, simple_loss=0.3406, pruned_loss=0.1053, over 1616493.70 frames. ], batch size: 21, lr: 1.23e-02, grad_scale: 4.0 +2023-02-06 03:06:12,492 INFO [optim.py:369] (0/4) 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 03:06:34,847 INFO [train.py:901] (0/4) Epoch 6, batch 6750, loss[loss=0.2407, simple_loss=0.3288, pruned_loss=0.07627, over 7825.00 frames. ], tot_loss[loss=0.2746, simple_loss=0.3398, pruned_loss=0.1047, over 1617070.06 frames. ], batch size: 20, lr: 1.23e-02, grad_scale: 4.0 +2023-02-06 03:06:38,953 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=47172.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:07:10,701 INFO [train.py:901] (0/4) Epoch 6, batch 6800, loss[loss=0.2781, simple_loss=0.3484, pruned_loss=0.1039, over 8102.00 frames. ], tot_loss[loss=0.2738, simple_loss=0.3389, pruned_loss=0.1043, over 1616068.64 frames. ], batch size: 23, lr: 1.23e-02, grad_scale: 8.0 +2023-02-06 03:07:12,684 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-06 03:07:21,197 INFO [optim.py:369] (0/4) 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 03:07:40,021 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=47259.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:07:44,574 INFO [train.py:901] (0/4) Epoch 6, batch 6850, loss[loss=0.2895, simple_loss=0.3324, pruned_loss=0.1233, over 7658.00 frames. ], tot_loss[loss=0.2747, simple_loss=0.3397, pruned_loss=0.1048, over 1615678.04 frames. ], batch size: 19, lr: 1.23e-02, grad_scale: 8.0 +2023-02-06 03:07:49,205 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.25 vs. limit=5.0 +2023-02-06 03:07:58,968 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47287.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:08:00,934 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-06 03:08:19,247 INFO [train.py:901] (0/4) Epoch 6, batch 6900, loss[loss=0.232, simple_loss=0.2992, pruned_loss=0.08242, over 7431.00 frames. ], tot_loss[loss=0.2766, simple_loss=0.3409, pruned_loss=0.1061, over 1613408.59 frames. ], batch size: 17, lr: 1.23e-02, grad_scale: 8.0 +2023-02-06 03:08:28,203 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47328.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:08:30,614 INFO [optim.py:369] (0/4) 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 03:08:32,864 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47335.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:08:44,849 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47353.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:08:49,769 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 6, batch 6950, loss[loss=0.2541, simple_loss=0.3294, pruned_loss=0.08938, over 8087.00 frames. ], tot_loss[loss=0.2761, simple_loss=0.3408, pruned_loss=0.1057, over 1613600.15 frames. ], batch size: 21, lr: 1.23e-02, grad_scale: 8.0 +2023-02-06 03:09:00,533 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3616, 1.7550, 1.7012, 0.9878, 1.7448, 1.3360, 0.3077, 1.5881], + device='cuda:0'), covar=tensor([0.0184, 0.0131, 0.0125, 0.0162, 0.0153, 0.0397, 0.0326, 0.0097], + device='cuda:0'), in_proj_covar=tensor([0.0333, 0.0250, 0.0207, 0.0300, 0.0243, 0.0390, 0.0309, 0.0285], + device='cuda:0'), out_proj_covar=tensor([1.0959e-04, 8.0680e-05, 6.5955e-05, 9.6708e-05, 7.9365e-05, 1.3656e-04, + 1.0196e-04, 9.2720e-05], device='cuda:0') +2023-02-06 03:09:09,773 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-06 03:09:15,140 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 6, batch 7000, loss[loss=0.2677, simple_loss=0.3431, pruned_loss=0.0961, over 8109.00 frames. ], tot_loss[loss=0.2734, simple_loss=0.3386, pruned_loss=0.1041, over 1615698.67 frames. ], batch size: 23, lr: 1.23e-02, grad_scale: 8.0 +2023-02-06 03:09:39,928 INFO [optim.py:369] (0/4) 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 03:10:03,570 INFO [train.py:901] (0/4) Epoch 6, batch 7050, loss[loss=0.2502, simple_loss=0.3283, pruned_loss=0.08607, over 8447.00 frames. ], tot_loss[loss=0.2726, simple_loss=0.3381, pruned_loss=0.1036, over 1618056.65 frames. ], batch size: 27, lr: 1.23e-02, grad_scale: 8.0 +2023-02-06 03:10:37,637 INFO [train.py:901] (0/4) Epoch 6, batch 7100, loss[loss=0.2708, simple_loss=0.3462, pruned_loss=0.09772, over 8461.00 frames. ], tot_loss[loss=0.2726, simple_loss=0.3383, pruned_loss=0.1034, over 1619802.39 frames. ], batch size: 27, lr: 1.22e-02, grad_scale: 8.0 +2023-02-06 03:10:48,827 INFO [optim.py:369] (0/4) 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 03:10:56,276 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47543.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:11:12,596 INFO [train.py:901] (0/4) Epoch 6, batch 7150, loss[loss=0.2067, simple_loss=0.3015, pruned_loss=0.05596, over 8315.00 frames. ], tot_loss[loss=0.2724, simple_loss=0.3383, pruned_loss=0.1032, over 1621530.55 frames. ], batch size: 25, lr: 1.22e-02, grad_scale: 8.0 +2023-02-06 03:11:14,075 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47568.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:11:22,813 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.22 vs. limit=5.0 +2023-02-06 03:11:37,932 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=47603.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:11:46,767 INFO [train.py:901] (0/4) Epoch 6, batch 7200, loss[loss=0.2746, simple_loss=0.3394, pruned_loss=0.1048, over 8080.00 frames. ], tot_loss[loss=0.2726, simple_loss=0.3387, pruned_loss=0.1033, over 1621992.54 frames. ], batch size: 21, lr: 1.22e-02, grad_scale: 8.0 +2023-02-06 03:11:49,876 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.50 vs. limit=5.0 +2023-02-06 03:11:57,784 INFO [optim.py:369] (0/4) 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 03:12:22,017 INFO [train.py:901] (0/4) Epoch 6, batch 7250, loss[loss=0.2599, simple_loss=0.3393, pruned_loss=0.09022, over 8504.00 frames. ], tot_loss[loss=0.2714, simple_loss=0.3382, pruned_loss=0.1023, over 1623778.66 frames. ], batch size: 28, lr: 1.22e-02, grad_scale: 8.0 +2023-02-06 03:12:56,472 INFO [train.py:901] (0/4) Epoch 6, batch 7300, loss[loss=0.2376, simple_loss=0.3053, pruned_loss=0.08489, over 8086.00 frames. ], tot_loss[loss=0.2723, simple_loss=0.3387, pruned_loss=0.1029, over 1622126.58 frames. ], batch size: 21, lr: 1.22e-02, grad_scale: 8.0 +2023-02-06 03:12:57,892 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47718.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:13:07,200 INFO [optim.py:369] (0/4) 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 03:13:13,254 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=47741.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:13:23,397 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=47756.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:13:30,044 INFO [train.py:901] (0/4) Epoch 6, batch 7350, loss[loss=0.2672, simple_loss=0.3354, pruned_loss=0.0995, over 7808.00 frames. ], tot_loss[loss=0.2719, simple_loss=0.3384, pruned_loss=0.1027, over 1619024.09 frames. ], batch size: 20, lr: 1.22e-02, grad_scale: 8.0 +2023-02-06 03:13:48,783 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-06 03:13:59,269 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=47806.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:14:05,860 INFO [train.py:901] (0/4) Epoch 6, batch 7400, loss[loss=0.2295, simple_loss=0.3054, pruned_loss=0.0768, over 7807.00 frames. ], tot_loss[loss=0.2728, simple_loss=0.339, pruned_loss=0.1034, over 1618648.16 frames. ], batch size: 20, lr: 1.22e-02, grad_scale: 8.0 +2023-02-06 03:14:08,028 WARNING [train.py:1067] (0/4) 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] (0/4) 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] (0/4) 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 03:14:33,710 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47856.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:14:40,236 INFO [train.py:901] (0/4) Epoch 6, batch 7450, loss[loss=0.2253, simple_loss=0.2922, pruned_loss=0.07917, over 7189.00 frames. ], tot_loss[loss=0.2721, simple_loss=0.3379, pruned_loss=0.1032, over 1612440.01 frames. ], batch size: 16, lr: 1.22e-02, grad_scale: 8.0 +2023-02-06 03:14:46,244 WARNING [train.py:1067] (0/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-06 03:15:15,310 INFO [train.py:901] (0/4) Epoch 6, batch 7500, loss[loss=0.2541, simple_loss=0.3403, pruned_loss=0.08389, over 8106.00 frames. ], tot_loss[loss=0.2717, simple_loss=0.3372, pruned_loss=0.1031, over 1610517.88 frames. ], batch size: 23, lr: 1.22e-02, grad_scale: 8.0 +2023-02-06 03:15:25,968 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 6, batch 7550, loss[loss=0.2665, simple_loss=0.3383, pruned_loss=0.0973, over 8473.00 frames. ], tot_loss[loss=0.271, simple_loss=0.3369, pruned_loss=0.1025, over 1610457.69 frames. ], batch size: 27, lr: 1.22e-02, grad_scale: 8.0 +2023-02-06 03:15:54,919 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47974.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:16:11,861 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47999.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:16:12,443 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-48000.pt +2023-02-06 03:16:24,750 INFO [train.py:901] (0/4) Epoch 6, batch 7600, loss[loss=0.3191, simple_loss=0.3718, pruned_loss=0.1332, over 7291.00 frames. ], tot_loss[loss=0.2729, simple_loss=0.3383, pruned_loss=0.1037, over 1613651.97 frames. ], batch size: 71, lr: 1.22e-02, grad_scale: 8.0 +2023-02-06 03:16:32,919 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.95 vs. limit=5.0 +2023-02-06 03:16:37,192 INFO [optim.py:369] (0/4) 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 03:17:01,515 INFO [train.py:901] (0/4) Epoch 6, batch 7650, loss[loss=0.2396, simple_loss=0.3142, pruned_loss=0.08254, over 8129.00 frames. ], tot_loss[loss=0.2726, simple_loss=0.3381, pruned_loss=0.1036, over 1612781.74 frames. ], batch size: 22, lr: 1.22e-02, grad_scale: 8.0 +2023-02-06 03:17:17,561 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48090.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:17:24,159 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=48100.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:17:32,445 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 6, batch 7700, loss[loss=0.3035, simple_loss=0.3618, pruned_loss=0.1226, over 8384.00 frames. ], tot_loss[loss=0.2712, simple_loss=0.3365, pruned_loss=0.103, over 1606514.98 frames. ], batch size: 49, lr: 1.22e-02, grad_scale: 8.0 +2023-02-06 03:17:46,047 INFO [optim.py:369] (0/4) 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 03:17:50,869 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48137.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:17:57,332 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-06 03:17:59,321 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=48150.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:18:10,114 INFO [train.py:901] (0/4) Epoch 6, batch 7750, loss[loss=0.2668, simple_loss=0.3268, pruned_loss=0.1034, over 7803.00 frames. ], tot_loss[loss=0.271, simple_loss=0.3364, pruned_loss=0.1029, over 1606680.93 frames. ], batch size: 19, lr: 1.22e-02, grad_scale: 8.0 +2023-02-06 03:18:13,410 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=48171.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 03:18:43,285 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 6, batch 7800, loss[loss=0.2958, simple_loss=0.3613, pruned_loss=0.1152, over 8786.00 frames. ], tot_loss[loss=0.2718, simple_loss=0.3371, pruned_loss=0.1032, over 1609728.45 frames. ], batch size: 30, lr: 1.22e-02, grad_scale: 8.0 +2023-02-06 03:18:53,435 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48230.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 03:18:54,592 INFO [optim.py:369] (0/4) 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 03:19:05,483 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3707, 2.6810, 1.6154, 2.0742, 2.0730, 1.3962, 1.8389, 2.1850], + device='cuda:0'), covar=tensor([0.1248, 0.0351, 0.1148, 0.0625, 0.0714, 0.1316, 0.1040, 0.0693], + device='cuda:0'), in_proj_covar=tensor([0.0343, 0.0229, 0.0314, 0.0301, 0.0312, 0.0314, 0.0338, 0.0312], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 03:19:10,040 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5959, 2.3908, 4.2812, 1.2125, 2.6374, 1.7911, 1.8581, 2.1960], + device='cuda:0'), covar=tensor([0.1696, 0.2038, 0.0780, 0.3704, 0.1787, 0.2978, 0.1622, 0.3014], + device='cuda:0'), in_proj_covar=tensor([0.0465, 0.0466, 0.0528, 0.0547, 0.0592, 0.0530, 0.0447, 0.0588], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003], + device='cuda:0') +2023-02-06 03:19:16,568 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48265.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:19:17,081 INFO [train.py:901] (0/4) Epoch 6, batch 7850, loss[loss=0.2822, simple_loss=0.349, pruned_loss=0.1078, over 8735.00 frames. ], tot_loss[loss=0.2718, simple_loss=0.337, pruned_loss=0.1033, over 1608638.08 frames. ], batch size: 30, lr: 1.22e-02, grad_scale: 8.0 +2023-02-06 03:19:30,611 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48286.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 03:19:35,338 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2935, 1.6530, 1.6407, 1.4488, 1.0501, 1.3908, 1.6776, 1.7592], + device='cuda:0'), covar=tensor([0.0486, 0.1130, 0.1653, 0.1266, 0.0604, 0.1446, 0.0709, 0.0534], + device='cuda:0'), in_proj_covar=tensor([0.0119, 0.0166, 0.0208, 0.0169, 0.0118, 0.0174, 0.0129, 0.0146], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0007, 0.0008, 0.0007, 0.0005, 0.0007, 0.0006, 0.0006], + device='cuda:0') +2023-02-06 03:19:51,006 INFO [train.py:901] (0/4) Epoch 6, batch 7900, loss[loss=0.2904, simple_loss=0.3555, pruned_loss=0.1126, over 8522.00 frames. ], tot_loss[loss=0.2711, simple_loss=0.3365, pruned_loss=0.1028, over 1611999.03 frames. ], batch size: 26, lr: 1.21e-02, grad_scale: 8.0 +2023-02-06 03:20:01,872 INFO [optim.py:369] (0/4) 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 03:20:09,333 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48342.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:20:10,797 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6641, 2.0439, 2.2664, 0.9644, 2.3874, 1.6441, 0.7134, 1.8950], + device='cuda:0'), covar=tensor([0.0312, 0.0147, 0.0137, 0.0290, 0.0160, 0.0367, 0.0421, 0.0152], + device='cuda:0'), in_proj_covar=tensor([0.0339, 0.0248, 0.0202, 0.0302, 0.0240, 0.0388, 0.0314, 0.0288], + device='cuda:0'), out_proj_covar=tensor([1.1131e-04, 7.9301e-05, 6.3516e-05, 9.6137e-05, 7.7835e-05, 1.3544e-04, + 1.0276e-04, 9.3068e-05], device='cuda:0') +2023-02-06 03:20:25,105 INFO [train.py:901] (0/4) Epoch 6, batch 7950, loss[loss=0.225, simple_loss=0.2956, pruned_loss=0.07721, over 7814.00 frames. ], tot_loss[loss=0.2718, simple_loss=0.3373, pruned_loss=0.1031, over 1612877.20 frames. ], batch size: 20, lr: 1.21e-02, grad_scale: 8.0 +2023-02-06 03:20:55,091 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.6213, 4.6592, 4.1135, 1.6717, 4.0020, 4.1267, 4.2356, 3.7302], + device='cuda:0'), covar=tensor([0.0819, 0.0604, 0.1176, 0.5280, 0.0846, 0.0728, 0.1321, 0.0812], + device='cuda:0'), in_proj_covar=tensor([0.0397, 0.0312, 0.0330, 0.0413, 0.0322, 0.0301, 0.0312, 0.0265], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 03:20:58,700 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48415.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:20:59,246 INFO [train.py:901] (0/4) Epoch 6, batch 8000, loss[loss=0.297, simple_loss=0.3396, pruned_loss=0.1272, over 7254.00 frames. ], tot_loss[loss=0.2725, simple_loss=0.3377, pruned_loss=0.1037, over 1609146.94 frames. ], batch size: 16, lr: 1.21e-02, grad_scale: 8.0 +2023-02-06 03:21:10,362 INFO [optim.py:369] (0/4) 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 03:21:11,761 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=48434.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:21:16,739 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48441.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:21:22,407 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 +2023-02-06 03:21:33,916 INFO [train.py:901] (0/4) Epoch 6, batch 8050, loss[loss=0.2376, simple_loss=0.3024, pruned_loss=0.08641, over 7538.00 frames. ], tot_loss[loss=0.2709, simple_loss=0.3353, pruned_loss=0.1032, over 1585261.79 frames. ], batch size: 18, lr: 1.21e-02, grad_scale: 8.0 +2023-02-06 03:21:37,653 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48471.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:21:54,558 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48496.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:21:56,889 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/epoch-6.pt +2023-02-06 03:22:07,495 WARNING [train.py:1067] (0/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-06 03:22:10,944 INFO [train.py:901] (0/4) Epoch 7, batch 0, loss[loss=0.239, simple_loss=0.3063, pruned_loss=0.08582, over 7809.00 frames. ], tot_loss[loss=0.239, simple_loss=0.3063, pruned_loss=0.08582, over 7809.00 frames. ], batch size: 19, lr: 1.14e-02, grad_scale: 8.0 +2023-02-06 03:22:10,945 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 03:22:22,759 INFO [train.py:935] (0/4) Epoch 7, validation: loss=0.2113, simple_loss=0.3091, pruned_loss=0.05678, over 944034.00 frames. +2023-02-06 03:22:22,760 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6689MB +2023-02-06 03:22:37,619 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48521.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:22:38,082 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-06 03:22:41,832 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-06 03:22:45,433 INFO [optim.py:369] (0/4) 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 03:22:46,419 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5294, 2.3259, 4.5193, 1.2988, 2.9451, 2.1243, 1.6920, 2.8667], + device='cuda:0'), covar=tensor([0.1516, 0.1785, 0.0564, 0.3259, 0.1461, 0.2365, 0.1481, 0.1959], + device='cuda:0'), in_proj_covar=tensor([0.0475, 0.0474, 0.0530, 0.0554, 0.0595, 0.0538, 0.0452, 0.0596], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003], + device='cuda:0') +2023-02-06 03:22:53,309 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48542.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 03:22:55,835 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48546.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:22:57,635 INFO [train.py:901] (0/4) Epoch 7, batch 50, loss[loss=0.3093, simple_loss=0.3709, pruned_loss=0.1238, over 8325.00 frames. ], tot_loss[loss=0.2682, simple_loss=0.3354, pruned_loss=0.1005, over 365673.56 frames. ], batch size: 26, lr: 1.14e-02, grad_scale: 8.0 +2023-02-06 03:22:57,793 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48549.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:23:09,795 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48567.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 03:23:12,925 WARNING [train.py:1067] (0/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-06 03:23:14,279 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=48574.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 03:23:22,854 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4835, 1.3626, 4.3203, 2.1321, 2.4082, 4.9274, 4.8792, 4.3951], + device='cuda:0'), covar=tensor([0.0998, 0.1465, 0.0227, 0.1710, 0.0888, 0.0228, 0.0320, 0.0553], + device='cuda:0'), in_proj_covar=tensor([0.0240, 0.0273, 0.0227, 0.0269, 0.0236, 0.0211, 0.0264, 0.0281], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 03:23:25,142 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.76 vs. limit=5.0 +2023-02-06 03:23:31,387 INFO [train.py:901] (0/4) Epoch 7, batch 100, loss[loss=0.2491, simple_loss=0.3244, pruned_loss=0.08685, over 7932.00 frames. ], tot_loss[loss=0.2705, simple_loss=0.3373, pruned_loss=0.1018, over 644102.63 frames. ], batch size: 20, lr: 1.14e-02, grad_scale: 8.0 +2023-02-06 03:23:34,997 WARNING [train.py:1067] (0/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-06 03:23:54,579 INFO [optim.py:369] (0/4) 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 03:24:06,736 INFO [train.py:901] (0/4) Epoch 7, batch 150, loss[loss=0.2529, simple_loss=0.3264, pruned_loss=0.08969, over 8653.00 frames. ], tot_loss[loss=0.2736, simple_loss=0.3393, pruned_loss=0.104, over 860109.51 frames. ], batch size: 34, lr: 1.13e-02, grad_scale: 8.0 +2023-02-06 03:24:08,457 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.65 vs. limit=5.0 +2023-02-06 03:24:09,542 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0897, 2.5829, 3.4182, 0.9914, 3.2929, 2.1220, 1.5824, 2.0057], + device='cuda:0'), covar=tensor([0.0359, 0.0167, 0.0106, 0.0349, 0.0193, 0.0361, 0.0394, 0.0209], + device='cuda:0'), in_proj_covar=tensor([0.0345, 0.0251, 0.0206, 0.0306, 0.0240, 0.0390, 0.0315, 0.0289], + device='cuda:0'), out_proj_covar=tensor([1.1297e-04, 8.0335e-05, 6.4819e-05, 9.7474e-05, 7.7261e-05, 1.3538e-04, + 1.0264e-04, 9.3195e-05], device='cuda:0') +2023-02-06 03:24:31,946 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=48686.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:24:34,067 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48689.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 03:24:40,629 INFO [train.py:901] (0/4) Epoch 7, batch 200, loss[loss=0.2925, simple_loss=0.3468, pruned_loss=0.1191, over 8139.00 frames. ], tot_loss[loss=0.2721, simple_loss=0.3382, pruned_loss=0.103, over 1026522.52 frames. ], batch size: 22, lr: 1.13e-02, grad_scale: 8.0 +2023-02-06 03:25:03,500 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 7, batch 250, loss[loss=0.2814, simple_loss=0.3586, pruned_loss=0.1021, over 8499.00 frames. ], tot_loss[loss=0.2709, simple_loss=0.3372, pruned_loss=0.1023, over 1156735.90 frames. ], batch size: 26, lr: 1.13e-02, grad_scale: 8.0 +2023-02-06 03:25:21,474 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2883, 1.1909, 2.2593, 1.1588, 2.0086, 2.4215, 2.4884, 2.0547], + device='cuda:0'), covar=tensor([0.0960, 0.1184, 0.0497, 0.1842, 0.0623, 0.0392, 0.0509, 0.0797], + device='cuda:0'), in_proj_covar=tensor([0.0237, 0.0271, 0.0226, 0.0267, 0.0235, 0.0211, 0.0263, 0.0277], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 03:25:22,702 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=48759.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:25:26,760 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-06 03:25:35,604 WARNING [train.py:1067] (0/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-06 03:25:41,107 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=48785.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:25:50,547 INFO [train.py:901] (0/4) Epoch 7, batch 300, loss[loss=0.2415, simple_loss=0.3284, pruned_loss=0.07728, over 8466.00 frames. ], tot_loss[loss=0.2717, simple_loss=0.3381, pruned_loss=0.1027, over 1259952.67 frames. ], batch size: 25, lr: 1.13e-02, grad_scale: 8.0 +2023-02-06 03:25:52,221 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48801.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:25:54,978 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48805.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:26:12,367 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48830.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:26:13,533 INFO [optim.py:369] (0/4) 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 03:26:18,388 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 7, batch 350, loss[loss=0.2518, simple_loss=0.3035, pruned_loss=0.1001, over 7440.00 frames. ], tot_loss[loss=0.2701, simple_loss=0.3368, pruned_loss=0.1017, over 1337161.43 frames. ], batch size: 17, lr: 1.13e-02, grad_scale: 8.0 +2023-02-06 03:26:30,603 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48856.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:26:43,218 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48874.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:27:00,371 INFO [train.py:901] (0/4) Epoch 7, batch 400, loss[loss=0.2512, simple_loss=0.3187, pruned_loss=0.09185, over 7796.00 frames. ], tot_loss[loss=0.2685, simple_loss=0.3353, pruned_loss=0.1008, over 1397314.46 frames. ], batch size: 19, lr: 1.13e-02, grad_scale: 8.0 +2023-02-06 03:27:01,276 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48900.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:27:09,181 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5519, 1.3910, 2.8455, 1.1690, 2.0443, 3.0587, 2.9937, 2.6734], + device='cuda:0'), covar=tensor([0.1000, 0.1337, 0.0417, 0.1961, 0.0767, 0.0283, 0.0487, 0.0636], + device='cuda:0'), in_proj_covar=tensor([0.0239, 0.0275, 0.0227, 0.0270, 0.0236, 0.0211, 0.0264, 0.0279], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 03:27:22,458 INFO [optim.py:369] (0/4) 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 03:27:32,152 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48945.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 03:27:34,584 INFO [train.py:901] (0/4) Epoch 7, batch 450, loss[loss=0.2992, simple_loss=0.3714, pruned_loss=0.1135, over 8105.00 frames. ], tot_loss[loss=0.27, simple_loss=0.3366, pruned_loss=0.1018, over 1444237.09 frames. ], batch size: 23, lr: 1.13e-02, grad_scale: 8.0 +2023-02-06 03:27:49,638 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48970.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 03:28:10,027 INFO [train.py:901] (0/4) Epoch 7, batch 500, loss[loss=0.3104, simple_loss=0.3667, pruned_loss=0.1271, over 8345.00 frames. ], tot_loss[loss=0.2694, simple_loss=0.3361, pruned_loss=0.1013, over 1481673.06 frames. ], batch size: 24, lr: 1.13e-02, grad_scale: 8.0 +2023-02-06 03:28:32,340 INFO [optim.py:369] (0/4) 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 03:28:33,279 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4225, 1.8454, 3.1405, 1.1727, 2.1679, 1.8434, 1.4285, 1.9612], + device='cuda:0'), covar=tensor([0.1581, 0.1822, 0.0633, 0.3359, 0.1417, 0.2558, 0.1664, 0.2136], + device='cuda:0'), in_proj_covar=tensor([0.0475, 0.0473, 0.0526, 0.0547, 0.0590, 0.0532, 0.0454, 0.0593], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:0') +2023-02-06 03:28:43,901 INFO [train.py:901] (0/4) Epoch 7, batch 550, loss[loss=0.2648, simple_loss=0.3286, pruned_loss=0.1005, over 8204.00 frames. ], tot_loss[loss=0.2675, simple_loss=0.3343, pruned_loss=0.1004, over 1507269.18 frames. ], batch size: 23, lr: 1.13e-02, grad_scale: 16.0 +2023-02-06 03:28:50,288 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49057.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:29:07,253 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49082.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:29:19,500 INFO [train.py:901] (0/4) Epoch 7, batch 600, loss[loss=0.3168, simple_loss=0.3625, pruned_loss=0.1355, over 6672.00 frames. ], tot_loss[loss=0.2671, simple_loss=0.3336, pruned_loss=0.1003, over 1529732.99 frames. ], batch size: 71, lr: 1.13e-02, grad_scale: 16.0 +2023-02-06 03:29:31,445 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-06 03:29:41,659 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49130.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:29:42,821 INFO [optim.py:369] (0/4) 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 03:29:54,594 INFO [train.py:901] (0/4) Epoch 7, batch 650, loss[loss=0.3089, simple_loss=0.374, pruned_loss=0.1219, over 8449.00 frames. ], tot_loss[loss=0.2655, simple_loss=0.3326, pruned_loss=0.09919, over 1549585.73 frames. ], batch size: 27, lr: 1.13e-02, grad_scale: 16.0 +2023-02-06 03:29:58,943 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49155.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:29:59,702 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49156.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:30:15,517 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.6066, 4.5378, 4.1288, 1.7576, 4.1333, 4.0504, 4.2674, 3.8420], + device='cuda:0'), covar=tensor([0.0697, 0.0573, 0.0825, 0.4284, 0.0687, 0.0707, 0.0951, 0.0613], + device='cuda:0'), in_proj_covar=tensor([0.0412, 0.0318, 0.0339, 0.0424, 0.0326, 0.0307, 0.0313, 0.0268], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 03:30:17,657 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49181.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:30:18,958 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=49183.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:30:28,558 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.9941, 1.2697, 4.2816, 1.5750, 3.6676, 3.5755, 3.7554, 3.6471], + device='cuda:0'), covar=tensor([0.0475, 0.3809, 0.0376, 0.2670, 0.1050, 0.0713, 0.0507, 0.0569], + device='cuda:0'), in_proj_covar=tensor([0.0371, 0.0515, 0.0458, 0.0454, 0.0518, 0.0428, 0.0432, 0.0487], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 03:30:29,824 INFO [train.py:901] (0/4) Epoch 7, batch 700, loss[loss=0.2302, simple_loss=0.2919, pruned_loss=0.08424, over 7548.00 frames. ], tot_loss[loss=0.265, simple_loss=0.3321, pruned_loss=0.09898, over 1561147.44 frames. ], batch size: 18, lr: 1.13e-02, grad_scale: 16.0 +2023-02-06 03:30:30,704 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=49200.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:30:54,554 INFO [optim.py:369] (0/4) 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 03:31:03,863 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-02-06 03:31:06,140 INFO [train.py:901] (0/4) Epoch 7, batch 750, loss[loss=0.2664, simple_loss=0.3291, pruned_loss=0.1018, over 7968.00 frames. ], tot_loss[loss=0.2666, simple_loss=0.3335, pruned_loss=0.09989, over 1573067.80 frames. ], batch size: 21, lr: 1.13e-02, grad_scale: 16.0 +2023-02-06 03:31:06,455 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.61 vs. limit=5.0 +2023-02-06 03:31:09,786 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49254.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:31:18,082 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-06 03:31:26,430 WARNING [train.py:1067] (0/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-06 03:31:40,949 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49298.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:31:41,433 INFO [train.py:901] (0/4) Epoch 7, batch 800, loss[loss=0.2944, simple_loss=0.3563, pruned_loss=0.1162, over 8320.00 frames. ], tot_loss[loss=0.2662, simple_loss=0.3333, pruned_loss=0.09955, over 1584927.33 frames. ], batch size: 26, lr: 1.13e-02, grad_scale: 16.0 +2023-02-06 03:31:52,694 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49315.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:31:59,464 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.4850, 4.5476, 4.0034, 2.0435, 4.0059, 4.1472, 4.2469, 3.8123], + device='cuda:0'), covar=tensor([0.0919, 0.0500, 0.0908, 0.4776, 0.0821, 0.0838, 0.1070, 0.0745], + device='cuda:0'), in_proj_covar=tensor([0.0413, 0.0319, 0.0346, 0.0430, 0.0331, 0.0312, 0.0318, 0.0270], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 03:32:05,537 INFO [optim.py:369] (0/4) 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 03:32:17,939 INFO [train.py:901] (0/4) Epoch 7, batch 850, loss[loss=0.2707, simple_loss=0.3476, pruned_loss=0.09687, over 8514.00 frames. ], tot_loss[loss=0.2665, simple_loss=0.3336, pruned_loss=0.09976, over 1593862.55 frames. ], batch size: 28, lr: 1.13e-02, grad_scale: 8.0 +2023-02-06 03:32:45,200 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3113, 1.7321, 2.7939, 1.1412, 1.8664, 1.6355, 1.4359, 1.6930], + device='cuda:0'), covar=tensor([0.1377, 0.1539, 0.0600, 0.2737, 0.1254, 0.2220, 0.1400, 0.1769], + device='cuda:0'), in_proj_covar=tensor([0.0471, 0.0471, 0.0529, 0.0545, 0.0590, 0.0527, 0.0450, 0.0593], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:0') +2023-02-06 03:32:52,553 INFO [train.py:901] (0/4) Epoch 7, batch 900, loss[loss=0.2512, simple_loss=0.3243, pruned_loss=0.08904, over 7970.00 frames. ], tot_loss[loss=0.2674, simple_loss=0.3344, pruned_loss=0.1002, over 1597202.88 frames. ], batch size: 21, lr: 1.13e-02, grad_scale: 8.0 +2023-02-06 03:33:17,134 INFO [optim.py:369] (0/4) 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 03:33:21,981 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49440.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:33:28,081 INFO [train.py:901] (0/4) Epoch 7, batch 950, loss[loss=0.2276, simple_loss=0.2981, pruned_loss=0.07853, over 7541.00 frames. ], tot_loss[loss=0.2681, simple_loss=0.335, pruned_loss=0.1006, over 1600050.18 frames. ], batch size: 18, lr: 1.13e-02, grad_scale: 8.0 +2023-02-06 03:33:28,868 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49450.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 03:33:50,514 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-06 03:33:50,653 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.4644, 1.0568, 4.6146, 1.6247, 3.9917, 3.7495, 4.1979, 4.0400], + device='cuda:0'), covar=tensor([0.0371, 0.4280, 0.0384, 0.2969, 0.1072, 0.0704, 0.0436, 0.0497], + device='cuda:0'), in_proj_covar=tensor([0.0369, 0.0516, 0.0464, 0.0460, 0.0522, 0.0432, 0.0438, 0.0489], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 03:34:03,650 INFO [train.py:901] (0/4) Epoch 7, batch 1000, loss[loss=0.2512, simple_loss=0.3197, pruned_loss=0.09139, over 7924.00 frames. ], tot_loss[loss=0.2655, simple_loss=0.3332, pruned_loss=0.09885, over 1604639.86 frames. ], batch size: 20, lr: 1.13e-02, grad_scale: 8.0 +2023-02-06 03:34:24,199 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-06 03:34:27,690 INFO [optim.py:369] (0/4) 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 03:34:35,910 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-06 03:34:38,740 INFO [train.py:901] (0/4) Epoch 7, batch 1050, loss[loss=0.2808, simple_loss=0.336, pruned_loss=0.1128, over 7927.00 frames. ], tot_loss[loss=0.2667, simple_loss=0.3344, pruned_loss=0.09951, over 1610682.75 frames. ], batch size: 20, lr: 1.12e-02, grad_scale: 8.0 +2023-02-06 03:34:43,039 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49554.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:34:54,535 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49571.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:35:00,737 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49579.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:35:13,191 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2115, 1.5963, 3.2945, 1.3825, 2.2487, 3.6816, 3.6297, 3.1194], + device='cuda:0'), covar=tensor([0.0842, 0.1391, 0.0388, 0.1902, 0.0833, 0.0231, 0.0400, 0.0589], + device='cuda:0'), in_proj_covar=tensor([0.0238, 0.0274, 0.0230, 0.0270, 0.0233, 0.0213, 0.0266, 0.0277], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 03:35:13,238 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49596.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:35:14,556 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=49598.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:35:15,146 INFO [train.py:901] (0/4) Epoch 7, batch 1100, loss[loss=0.3226, simple_loss=0.3754, pruned_loss=0.1349, over 7045.00 frames. ], tot_loss[loss=0.2672, simple_loss=0.3349, pruned_loss=0.09982, over 1610794.15 frames. ], batch size: 72, lr: 1.12e-02, grad_scale: 8.0 +2023-02-06 03:35:38,352 INFO [optim.py:369] (0/4) 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 03:35:46,770 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-06 03:35:49,452 INFO [train.py:901] (0/4) Epoch 7, batch 1150, loss[loss=0.2595, simple_loss=0.3288, pruned_loss=0.09506, over 8445.00 frames. ], tot_loss[loss=0.2672, simple_loss=0.3346, pruned_loss=0.09988, over 1614689.72 frames. ], batch size: 29, lr: 1.12e-02, grad_scale: 8.0 +2023-02-06 03:36:23,378 INFO [train.py:901] (0/4) Epoch 7, batch 1200, loss[loss=0.2573, simple_loss=0.3338, pruned_loss=0.09037, over 8633.00 frames. ], tot_loss[loss=0.2683, simple_loss=0.335, pruned_loss=0.1008, over 1610406.32 frames. ], batch size: 34, lr: 1.12e-02, grad_scale: 8.0 +2023-02-06 03:36:27,559 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8230, 1.4609, 1.4526, 1.2899, 1.0319, 1.2855, 1.4207, 1.3492], + device='cuda:0'), covar=tensor([0.0551, 0.1259, 0.1735, 0.1376, 0.0650, 0.1488, 0.0756, 0.0586], + device='cuda:0'), in_proj_covar=tensor([0.0118, 0.0167, 0.0206, 0.0170, 0.0117, 0.0175, 0.0129, 0.0146], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0007, 0.0008, 0.0007, 0.0005, 0.0007, 0.0006, 0.0006], + device='cuda:0') +2023-02-06 03:36:33,700 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49713.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:36:47,165 INFO [optim.py:369] (0/4) 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 03:36:57,971 INFO [train.py:901] (0/4) Epoch 7, batch 1250, loss[loss=0.2769, simple_loss=0.3324, pruned_loss=0.1107, over 7538.00 frames. ], tot_loss[loss=0.2689, simple_loss=0.3353, pruned_loss=0.1012, over 1613515.69 frames. ], batch size: 18, lr: 1.12e-02, grad_scale: 8.0 +2023-02-06 03:37:22,193 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=49784.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:37:22,269 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49784.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:37:29,683 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=49794.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 03:37:32,814 INFO [train.py:901] (0/4) Epoch 7, batch 1300, loss[loss=0.2973, simple_loss=0.3572, pruned_loss=0.1187, over 8132.00 frames. ], tot_loss[loss=0.2677, simple_loss=0.3349, pruned_loss=0.1002, over 1616519.38 frames. ], batch size: 22, lr: 1.12e-02, grad_scale: 8.0 +2023-02-06 03:37:57,661 INFO [optim.py:369] (0/4) 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 03:38:08,125 INFO [train.py:901] (0/4) Epoch 7, batch 1350, loss[loss=0.2587, simple_loss=0.3272, pruned_loss=0.0951, over 8025.00 frames. ], tot_loss[loss=0.2689, simple_loss=0.3358, pruned_loss=0.1009, over 1618054.59 frames. ], batch size: 22, lr: 1.12e-02, grad_scale: 8.0 +2023-02-06 03:38:42,144 INFO [train.py:901] (0/4) Epoch 7, batch 1400, loss[loss=0.2765, simple_loss=0.3484, pruned_loss=0.1023, over 8286.00 frames. ], tot_loss[loss=0.269, simple_loss=0.3361, pruned_loss=0.101, over 1616250.59 frames. ], batch size: 23, lr: 1.12e-02, grad_scale: 8.0 +2023-02-06 03:38:42,332 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49899.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:38:49,798 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49909.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 03:38:53,179 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 +2023-02-06 03:39:07,157 INFO [optim.py:369] (0/4) 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 03:39:11,355 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-06 03:39:18,097 INFO [train.py:901] (0/4) Epoch 7, batch 1450, loss[loss=0.2393, simple_loss=0.301, pruned_loss=0.08884, over 7258.00 frames. ], tot_loss[loss=0.2682, simple_loss=0.3351, pruned_loss=0.1006, over 1617733.32 frames. ], batch size: 16, lr: 1.12e-02, grad_scale: 8.0 +2023-02-06 03:39:26,259 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3101, 1.5495, 4.4363, 1.8882, 2.3785, 5.1012, 4.9041, 4.4654], + device='cuda:0'), covar=tensor([0.1009, 0.1531, 0.0260, 0.1919, 0.0915, 0.0156, 0.0304, 0.0451], + device='cuda:0'), in_proj_covar=tensor([0.0238, 0.0275, 0.0231, 0.0272, 0.0236, 0.0215, 0.0269, 0.0281], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 03:39:31,675 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49969.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:39:41,668 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.45 vs. limit=5.0 +2023-02-06 03:39:49,073 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49994.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:39:52,393 INFO [train.py:901] (0/4) Epoch 7, batch 1500, loss[loss=0.2594, simple_loss=0.3348, pruned_loss=0.09198, over 8344.00 frames. ], tot_loss[loss=0.2672, simple_loss=0.3345, pruned_loss=0.09999, over 1615232.47 frames. ], batch size: 25, lr: 1.12e-02, grad_scale: 8.0 +2023-02-06 03:39:53,213 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-50000.pt +2023-02-06 03:40:07,868 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 +2023-02-06 03:40:16,579 INFO [optim.py:369] (0/4) 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 03:40:27,903 INFO [train.py:901] (0/4) Epoch 7, batch 1550, loss[loss=0.3368, simple_loss=0.3925, pruned_loss=0.1406, over 8504.00 frames. ], tot_loss[loss=0.267, simple_loss=0.334, pruned_loss=0.1, over 1610298.12 frames. ], batch size: 26, lr: 1.12e-02, grad_scale: 8.0 +2023-02-06 03:40:57,943 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5216, 1.7083, 1.9363, 1.6301, 1.0284, 1.8787, 0.2065, 1.1562], + device='cuda:0'), covar=tensor([0.3115, 0.1985, 0.0883, 0.1931, 0.6052, 0.0579, 0.4434, 0.2336], + device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0134, 0.0082, 0.0186, 0.0225, 0.0084, 0.0144, 0.0140], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 03:41:02,316 INFO [train.py:901] (0/4) Epoch 7, batch 1600, loss[loss=0.3186, simple_loss=0.3712, pruned_loss=0.133, over 8081.00 frames. ], tot_loss[loss=0.2672, simple_loss=0.334, pruned_loss=0.1002, over 1606781.55 frames. ], batch size: 21, lr: 1.12e-02, grad_scale: 8.0 +2023-02-06 03:41:22,664 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=50128.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:41:25,922 INFO [optim.py:369] (0/4) 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 03:41:35,372 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50146.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:41:35,569 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 +2023-02-06 03:41:37,165 INFO [train.py:901] (0/4) Epoch 7, batch 1650, loss[loss=0.218, simple_loss=0.2922, pruned_loss=0.0719, over 7794.00 frames. ], tot_loss[loss=0.2667, simple_loss=0.3338, pruned_loss=0.09987, over 1608911.84 frames. ], batch size: 19, lr: 1.12e-02, grad_scale: 8.0 +2023-02-06 03:41:40,091 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9526, 2.7200, 3.1553, 1.0945, 3.1726, 1.9486, 1.5164, 2.0188], + device='cuda:0'), covar=tensor([0.0416, 0.0150, 0.0110, 0.0362, 0.0147, 0.0389, 0.0413, 0.0217], + device='cuda:0'), in_proj_covar=tensor([0.0348, 0.0258, 0.0215, 0.0311, 0.0249, 0.0403, 0.0315, 0.0292], + device='cuda:0'), out_proj_covar=tensor([1.1342e-04, 8.1425e-05, 6.7136e-05, 9.8328e-05, 7.9676e-05, 1.3926e-04, + 1.0242e-04, 9.3340e-05], device='cuda:0') +2023-02-06 03:41:41,427 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50155.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:41:48,662 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50165.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 03:41:59,244 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50180.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:42:05,842 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50190.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 03:42:11,953 INFO [train.py:901] (0/4) Epoch 7, batch 1700, loss[loss=0.2291, simple_loss=0.3015, pruned_loss=0.07831, over 8083.00 frames. ], tot_loss[loss=0.266, simple_loss=0.333, pruned_loss=0.0995, over 1607717.39 frames. ], batch size: 21, lr: 1.12e-02, grad_scale: 8.0 +2023-02-06 03:42:22,978 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 03:42:41,484 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9063, 2.4454, 4.7890, 1.3456, 3.1385, 2.2195, 2.0093, 2.7117], + device='cuda:0'), covar=tensor([0.1310, 0.1646, 0.0477, 0.2869, 0.1054, 0.2133, 0.1168, 0.1941], + device='cuda:0'), in_proj_covar=tensor([0.0479, 0.0472, 0.0530, 0.0550, 0.0590, 0.0534, 0.0451, 0.0592], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:0') +2023-02-06 03:42:42,135 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 7, batch 1750, loss[loss=0.3058, simple_loss=0.3727, pruned_loss=0.1194, over 8105.00 frames. ], tot_loss[loss=0.2665, simple_loss=0.3336, pruned_loss=0.09969, over 1609107.18 frames. ], batch size: 23, lr: 1.12e-02, grad_scale: 8.0 +2023-02-06 03:43:21,551 INFO [train.py:901] (0/4) Epoch 7, batch 1800, loss[loss=0.3123, simple_loss=0.3609, pruned_loss=0.1318, over 8123.00 frames. ], tot_loss[loss=0.2665, simple_loss=0.3335, pruned_loss=0.09977, over 1607572.18 frames. ], batch size: 22, lr: 1.12e-02, grad_scale: 8.0 +2023-02-06 03:43:44,791 INFO [optim.py:369] (0/4) 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 03:43:56,163 INFO [train.py:901] (0/4) Epoch 7, batch 1850, loss[loss=0.2528, simple_loss=0.3226, pruned_loss=0.0915, over 7800.00 frames. ], tot_loss[loss=0.2646, simple_loss=0.3318, pruned_loss=0.09869, over 1605144.41 frames. ], batch size: 19, lr: 1.12e-02, grad_scale: 8.0 +2023-02-06 03:44:23,780 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-06 03:44:28,232 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4298, 1.4792, 1.6323, 1.2666, 0.9286, 1.6907, 0.0838, 0.9870], + device='cuda:0'), covar=tensor([0.3321, 0.2010, 0.0926, 0.2176, 0.5201, 0.0616, 0.4158, 0.2231], + device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0138, 0.0082, 0.0188, 0.0226, 0.0086, 0.0145, 0.0140], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 03:44:30,595 INFO [train.py:901] (0/4) Epoch 7, batch 1900, loss[loss=0.2537, simple_loss=0.3254, pruned_loss=0.09098, over 8601.00 frames. ], tot_loss[loss=0.2659, simple_loss=0.333, pruned_loss=0.0994, over 1604849.89 frames. ], batch size: 34, lr: 1.12e-02, grad_scale: 8.0 +2023-02-06 03:44:43,989 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-06 03:44:53,920 INFO [optim.py:369] (0/4) 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 03:44:56,019 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-06 03:45:04,592 INFO [train.py:901] (0/4) Epoch 7, batch 1950, loss[loss=0.2894, simple_loss=0.335, pruned_loss=0.1219, over 7715.00 frames. ], tot_loss[loss=0.2679, simple_loss=0.3349, pruned_loss=0.1004, over 1610462.00 frames. ], batch size: 18, lr: 1.11e-02, grad_scale: 8.0 +2023-02-06 03:45:15,309 WARNING [train.py:1067] (0/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-06 03:45:33,289 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=50490.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:45:39,200 INFO [train.py:901] (0/4) Epoch 7, batch 2000, loss[loss=0.2586, simple_loss=0.3307, pruned_loss=0.09326, over 8583.00 frames. ], tot_loss[loss=0.2681, simple_loss=0.3355, pruned_loss=0.1004, over 1615003.14 frames. ], batch size: 34, lr: 1.11e-02, grad_scale: 8.0 +2023-02-06 03:45:39,410 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50499.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:45:56,773 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50524.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:46:03,226 INFO [optim.py:369] (0/4) 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 03:46:13,948 INFO [train.py:901] (0/4) Epoch 7, batch 2050, loss[loss=0.1986, simple_loss=0.2692, pruned_loss=0.06401, over 7545.00 frames. ], tot_loss[loss=0.2678, simple_loss=0.3355, pruned_loss=0.1, over 1619515.89 frames. ], batch size: 18, lr: 1.11e-02, grad_scale: 8.0 +2023-02-06 03:46:19,889 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.7501, 5.7721, 5.0614, 2.1434, 5.2280, 5.3996, 5.3205, 5.1106], + device='cuda:0'), covar=tensor([0.0489, 0.0372, 0.0823, 0.4518, 0.0638, 0.0506, 0.0948, 0.0432], + device='cuda:0'), in_proj_covar=tensor([0.0418, 0.0314, 0.0344, 0.0430, 0.0333, 0.0317, 0.0322, 0.0275], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 03:46:20,540 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=50559.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 03:46:23,780 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50564.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:46:47,297 INFO [train.py:901] (0/4) Epoch 7, batch 2100, loss[loss=0.3224, simple_loss=0.3723, pruned_loss=0.1363, over 8440.00 frames. ], tot_loss[loss=0.2684, simple_loss=0.3357, pruned_loss=0.1006, over 1619653.74 frames. ], batch size: 27, lr: 1.11e-02, grad_scale: 8.0 +2023-02-06 03:46:51,696 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50605.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:47:10,012 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.80 vs. limit=5.0 +2023-02-06 03:47:11,032 INFO [optim.py:369] (0/4) 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 03:47:22,377 INFO [train.py:901] (0/4) Epoch 7, batch 2150, loss[loss=0.3087, simple_loss=0.3677, pruned_loss=0.1249, over 8416.00 frames. ], tot_loss[loss=0.2667, simple_loss=0.3344, pruned_loss=0.09954, over 1619953.06 frames. ], batch size: 49, lr: 1.11e-02, grad_scale: 8.0 +2023-02-06 03:47:40,232 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50674.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 03:47:56,606 INFO [train.py:901] (0/4) Epoch 7, batch 2200, loss[loss=0.256, simple_loss=0.3264, pruned_loss=0.09279, over 8247.00 frames. ], tot_loss[loss=0.2674, simple_loss=0.3348, pruned_loss=0.09995, over 1618977.28 frames. ], batch size: 24, lr: 1.11e-02, grad_scale: 8.0 +2023-02-06 03:48:07,076 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.94 vs. limit=5.0 +2023-02-06 03:48:20,831 INFO [optim.py:369] (0/4) 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 03:48:31,245 INFO [train.py:901] (0/4) Epoch 7, batch 2250, loss[loss=0.2997, simple_loss=0.3636, pruned_loss=0.1179, over 8204.00 frames. ], tot_loss[loss=0.2679, simple_loss=0.3353, pruned_loss=0.1002, over 1619856.62 frames. ], batch size: 23, lr: 1.11e-02, grad_scale: 8.0 +2023-02-06 03:48:53,491 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 +2023-02-06 03:49:05,384 INFO [train.py:901] (0/4) Epoch 7, batch 2300, loss[loss=0.2819, simple_loss=0.3383, pruned_loss=0.1128, over 8460.00 frames. ], tot_loss[loss=0.2692, simple_loss=0.3361, pruned_loss=0.1011, over 1617232.39 frames. ], batch size: 48, lr: 1.11e-02, grad_scale: 8.0 +2023-02-06 03:49:06,296 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50800.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:49:23,395 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 03:49:30,326 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 +2023-02-06 03:49:39,826 INFO [train.py:901] (0/4) Epoch 7, batch 2350, loss[loss=0.2285, simple_loss=0.3005, pruned_loss=0.07823, over 7968.00 frames. ], tot_loss[loss=0.27, simple_loss=0.3363, pruned_loss=0.1018, over 1620645.71 frames. ], batch size: 21, lr: 1.11e-02, grad_scale: 8.0 +2023-02-06 03:49:47,967 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50861.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:50:04,996 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50886.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:50:14,222 INFO [train.py:901] (0/4) Epoch 7, batch 2400, loss[loss=0.2765, simple_loss=0.3478, pruned_loss=0.1026, over 8195.00 frames. ], tot_loss[loss=0.2679, simple_loss=0.3345, pruned_loss=0.1006, over 1619742.81 frames. ], batch size: 23, lr: 1.11e-02, grad_scale: 8.0 +2023-02-06 03:50:20,197 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=50908.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:50:32,392 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.7577, 3.6728, 3.3735, 1.7068, 3.2810, 3.2639, 3.4186, 2.8638], + device='cuda:0'), covar=tensor([0.0987, 0.0794, 0.1172, 0.4862, 0.0892, 0.0990, 0.1264, 0.0948], + device='cuda:0'), in_proj_covar=tensor([0.0417, 0.0316, 0.0350, 0.0436, 0.0337, 0.0313, 0.0320, 0.0275], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 03:50:35,192 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) Epoch 7, batch 2450, loss[loss=0.2364, simple_loss=0.3217, pruned_loss=0.07553, over 8294.00 frames. ], tot_loss[loss=0.2671, simple_loss=0.3339, pruned_loss=0.1002, over 1615387.36 frames. ], batch size: 23, lr: 1.11e-02, grad_scale: 8.0 +2023-02-06 03:50:52,346 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50955.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 03:50:59,071 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 7, batch 2500, loss[loss=0.2998, simple_loss=0.3652, pruned_loss=0.1172, over 8338.00 frames. ], tot_loss[loss=0.2665, simple_loss=0.3336, pruned_loss=0.09974, over 1617057.76 frames. ], batch size: 26, lr: 1.11e-02, grad_scale: 8.0 +2023-02-06 03:51:39,931 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51023.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:51:46,387 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 7, batch 2550, loss[loss=0.2649, simple_loss=0.3331, pruned_loss=0.09833, over 8032.00 frames. ], tot_loss[loss=0.2662, simple_loss=0.3334, pruned_loss=0.09949, over 1615723.52 frames. ], batch size: 22, lr: 1.11e-02, grad_scale: 8.0 +2023-02-06 03:52:16,208 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.2148, 3.0668, 2.9057, 1.3996, 2.8034, 2.8625, 2.9370, 2.6176], + device='cuda:0'), covar=tensor([0.1184, 0.0860, 0.1240, 0.4544, 0.1046, 0.1104, 0.1489, 0.1202], + device='cuda:0'), in_proj_covar=tensor([0.0404, 0.0308, 0.0336, 0.0422, 0.0326, 0.0305, 0.0315, 0.0269], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 03:52:18,133 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51080.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:52:31,078 INFO [train.py:901] (0/4) Epoch 7, batch 2600, loss[loss=0.3184, simple_loss=0.3783, pruned_loss=0.1293, over 8510.00 frames. ], tot_loss[loss=0.2672, simple_loss=0.3341, pruned_loss=0.1002, over 1618397.44 frames. ], batch size: 26, lr: 1.11e-02, grad_scale: 8.0 +2023-02-06 03:52:54,868 INFO [optim.py:369] (0/4) 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] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=51144.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:53:05,173 INFO [train.py:901] (0/4) Epoch 7, batch 2650, loss[loss=0.288, simple_loss=0.3544, pruned_loss=0.1108, over 8249.00 frames. ], tot_loss[loss=0.2673, simple_loss=0.3341, pruned_loss=0.1002, over 1615963.33 frames. ], batch size: 24, lr: 1.11e-02, grad_scale: 8.0 +2023-02-06 03:53:19,234 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=51170.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 03:53:32,640 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 7, batch 2700, loss[loss=0.2744, simple_loss=0.3237, pruned_loss=0.1126, over 7529.00 frames. ], tot_loss[loss=0.2674, simple_loss=0.3345, pruned_loss=0.1001, over 1613806.37 frames. ], batch size: 18, lr: 1.11e-02, grad_scale: 8.0 +2023-02-06 03:53:55,361 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4402, 1.7810, 1.8314, 0.9987, 1.9124, 1.2819, 0.5562, 1.5863], + device='cuda:0'), covar=tensor([0.0257, 0.0133, 0.0103, 0.0212, 0.0153, 0.0348, 0.0310, 0.0114], + device='cuda:0'), in_proj_covar=tensor([0.0344, 0.0252, 0.0215, 0.0308, 0.0251, 0.0392, 0.0309, 0.0288], + device='cuda:0'), out_proj_covar=tensor([1.1117e-04, 7.9022e-05, 6.6944e-05, 9.6476e-05, 7.9891e-05, 1.3440e-04, + 9.9883e-05, 9.1433e-05], device='cuda:0') +2023-02-06 03:54:02,460 INFO [optim.py:369] (0/4) 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 03:54:14,245 INFO [train.py:901] (0/4) Epoch 7, batch 2750, loss[loss=0.3025, simple_loss=0.346, pruned_loss=0.1295, over 7145.00 frames. ], tot_loss[loss=0.2684, simple_loss=0.3353, pruned_loss=0.1008, over 1616643.17 frames. ], batch size: 71, lr: 1.11e-02, grad_scale: 8.0 +2023-02-06 03:54:21,068 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51259.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:54:34,380 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51279.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:54:36,584 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.09 vs. limit=5.0 +2023-02-06 03:54:38,278 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51285.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 03:54:47,532 INFO [train.py:901] (0/4) Epoch 7, batch 2800, loss[loss=0.2911, simple_loss=0.3398, pruned_loss=0.1212, over 8336.00 frames. ], tot_loss[loss=0.2688, simple_loss=0.3352, pruned_loss=0.1012, over 1612470.58 frames. ], batch size: 26, lr: 1.11e-02, grad_scale: 8.0 +2023-02-06 03:54:51,134 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51304.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:54:53,733 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 03:55:22,689 INFO [train.py:901] (0/4) Epoch 7, batch 2850, loss[loss=0.3165, simple_loss=0.3689, pruned_loss=0.132, over 6924.00 frames. ], tot_loss[loss=0.2682, simple_loss=0.3352, pruned_loss=0.1006, over 1618337.56 frames. ], batch size: 71, lr: 1.10e-02, grad_scale: 16.0 +2023-02-06 03:55:48,728 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4897, 2.0620, 3.3633, 2.4679, 2.6822, 2.0658, 1.6336, 1.3766], + device='cuda:0'), covar=tensor([0.2611, 0.2721, 0.0620, 0.1865, 0.1539, 0.1613, 0.1490, 0.3120], + device='cuda:0'), in_proj_covar=tensor([0.0817, 0.0756, 0.0657, 0.0753, 0.0843, 0.0698, 0.0656, 0.0689], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 03:55:57,257 INFO [train.py:901] (0/4) Epoch 7, batch 2900, loss[loss=0.263, simple_loss=0.3188, pruned_loss=0.1036, over 7441.00 frames. ], tot_loss[loss=0.2666, simple_loss=0.3343, pruned_loss=0.09945, over 1617790.22 frames. ], batch size: 17, lr: 1.10e-02, grad_scale: 16.0 +2023-02-06 03:55:58,208 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51400.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:56:04,949 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51410.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:56:13,550 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=51424.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:56:19,633 WARNING [train.py:1067] (0/4) 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] (0/4) 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 03:56:31,120 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2351, 1.6177, 1.6172, 1.4040, 1.1878, 1.6002, 1.9210, 1.6871], + device='cuda:0'), covar=tensor([0.0526, 0.1154, 0.1732, 0.1383, 0.0637, 0.1453, 0.0639, 0.0572], + device='cuda:0'), in_proj_covar=tensor([0.0118, 0.0166, 0.0208, 0.0173, 0.0117, 0.0174, 0.0127, 0.0147], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0007, 0.0008, 0.0008, 0.0005, 0.0008, 0.0006, 0.0006], + device='cuda:0') +2023-02-06 03:56:32,317 INFO [train.py:901] (0/4) Epoch 7, batch 2950, loss[loss=0.234, simple_loss=0.2941, pruned_loss=0.08696, over 7221.00 frames. ], tot_loss[loss=0.2668, simple_loss=0.3343, pruned_loss=0.09966, over 1615373.16 frames. ], batch size: 16, lr: 1.10e-02, grad_scale: 16.0 +2023-02-06 03:57:06,418 INFO [train.py:901] (0/4) Epoch 7, batch 3000, loss[loss=0.2744, simple_loss=0.3473, pruned_loss=0.1008, over 8363.00 frames. ], tot_loss[loss=0.2671, simple_loss=0.3346, pruned_loss=0.09978, over 1617183.85 frames. ], batch size: 24, lr: 1.10e-02, grad_scale: 8.0 +2023-02-06 03:57:06,419 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 03:57:21,702 INFO [train.py:935] (0/4) Epoch 7, validation: loss=0.2071, simple_loss=0.305, pruned_loss=0.05459, over 944034.00 frames. +2023-02-06 03:57:21,703 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6689MB +2023-02-06 03:57:31,180 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51513.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:57:32,555 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51515.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:57:45,147 INFO [optim.py:369] (0/4) 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 03:57:45,235 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=51534.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 03:57:48,652 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51539.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:57:49,344 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51541.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 03:57:55,346 INFO [train.py:901] (0/4) Epoch 7, batch 3050, loss[loss=0.2536, simple_loss=0.3277, pruned_loss=0.08976, over 8330.00 frames. ], tot_loss[loss=0.267, simple_loss=0.3347, pruned_loss=0.09968, over 1614059.56 frames. ], batch size: 25, lr: 1.10e-02, grad_scale: 8.0 +2023-02-06 03:58:06,941 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51566.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 03:58:22,439 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 +2023-02-06 03:58:29,894 INFO [train.py:901] (0/4) Epoch 7, batch 3100, loss[loss=0.2652, simple_loss=0.3367, pruned_loss=0.0968, over 8514.00 frames. ], tot_loss[loss=0.2681, simple_loss=0.3352, pruned_loss=0.1004, over 1614444.75 frames. ], batch size: 28, lr: 1.10e-02, grad_scale: 8.0 +2023-02-06 03:58:54,843 INFO [optim.py:369] (0/4) 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 03:59:05,324 INFO [train.py:901] (0/4) Epoch 7, batch 3150, loss[loss=0.2357, simple_loss=0.3001, pruned_loss=0.08564, over 8083.00 frames. ], tot_loss[loss=0.2677, simple_loss=0.3343, pruned_loss=0.1005, over 1609967.25 frames. ], batch size: 21, lr: 1.10e-02, grad_scale: 8.0 +2023-02-06 03:59:05,482 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0266, 1.5170, 3.4969, 1.4725, 2.4071, 3.9637, 3.8461, 3.3866], + device='cuda:0'), covar=tensor([0.0974, 0.1330, 0.0299, 0.1721, 0.0676, 0.0201, 0.0382, 0.0560], + device='cuda:0'), in_proj_covar=tensor([0.0239, 0.0272, 0.0228, 0.0271, 0.0241, 0.0214, 0.0271, 0.0279], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 03:59:05,508 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51649.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 03:59:17,186 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.8087, 1.2741, 5.8965, 1.9959, 5.1087, 4.8137, 5.3907, 5.2832], + device='cuda:0'), covar=tensor([0.0368, 0.4247, 0.0280, 0.2894, 0.1008, 0.0709, 0.0365, 0.0422], + device='cuda:0'), in_proj_covar=tensor([0.0377, 0.0518, 0.0471, 0.0461, 0.0526, 0.0439, 0.0433, 0.0491], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 03:59:22,855 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4489, 1.8726, 3.1164, 1.2330, 2.1358, 1.7321, 1.6080, 1.8529], + device='cuda:0'), covar=tensor([0.1546, 0.1823, 0.0628, 0.3300, 0.1405, 0.2444, 0.1497, 0.2094], + device='cuda:0'), in_proj_covar=tensor([0.0473, 0.0471, 0.0530, 0.0549, 0.0588, 0.0528, 0.0451, 0.0589], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:0') +2023-02-06 03:59:26,399 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51679.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 03:59:40,118 INFO [train.py:901] (0/4) Epoch 7, batch 3200, loss[loss=0.2674, simple_loss=0.3338, pruned_loss=0.1006, over 7648.00 frames. ], tot_loss[loss=0.2659, simple_loss=0.3325, pruned_loss=0.09965, over 1608819.70 frames. ], batch size: 19, lr: 1.10e-02, grad_scale: 8.0 +2023-02-06 03:59:43,597 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51704.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:00:01,508 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0550, 1.2753, 1.1551, 0.4064, 1.2144, 0.9354, 0.0977, 1.1489], + device='cuda:0'), covar=tensor([0.0226, 0.0153, 0.0144, 0.0261, 0.0168, 0.0439, 0.0348, 0.0144], + device='cuda:0'), in_proj_covar=tensor([0.0353, 0.0260, 0.0217, 0.0317, 0.0255, 0.0404, 0.0316, 0.0294], + device='cuda:0'), out_proj_covar=tensor([1.1391e-04, 8.1587e-05, 6.7352e-05, 9.9332e-05, 8.0862e-05, 1.3819e-04, + 1.0200e-04, 9.2874e-05], device='cuda:0') +2023-02-06 04:00:05,271 INFO [optim.py:369] (0/4) 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 04:00:10,372 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 +2023-02-06 04:00:12,045 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=51744.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:00:16,038 INFO [train.py:901] (0/4) Epoch 7, batch 3250, loss[loss=0.2341, simple_loss=0.3128, pruned_loss=0.07766, over 8504.00 frames. ], tot_loss[loss=0.2653, simple_loss=0.3325, pruned_loss=0.09909, over 1614438.84 frames. ], batch size: 26, lr: 1.10e-02, grad_scale: 8.0 +2023-02-06 04:00:19,459 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=51754.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:00:26,774 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51765.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:00:46,789 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51795.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:00:49,330 INFO [train.py:901] (0/4) Epoch 7, batch 3300, loss[loss=0.2953, simple_loss=0.3599, pruned_loss=0.1153, over 8249.00 frames. ], tot_loss[loss=0.2648, simple_loss=0.3322, pruned_loss=0.0987, over 1611770.38 frames. ], batch size: 24, lr: 1.10e-02, grad_scale: 8.0 +2023-02-06 04:01:02,344 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9763, 2.6320, 3.1195, 1.1131, 3.1377, 1.8543, 1.4614, 1.9868], + device='cuda:0'), covar=tensor([0.0427, 0.0160, 0.0114, 0.0380, 0.0205, 0.0419, 0.0483, 0.0219], + device='cuda:0'), in_proj_covar=tensor([0.0347, 0.0253, 0.0213, 0.0312, 0.0249, 0.0397, 0.0309, 0.0288], + device='cuda:0'), out_proj_covar=tensor([1.1177e-04, 7.9248e-05, 6.6127e-05, 9.7606e-05, 7.8758e-05, 1.3592e-04, + 9.9669e-05, 9.0739e-05], device='cuda:0') +2023-02-06 04:01:03,663 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51820.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:01:14,385 INFO [optim.py:369] (0/4) 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 04:01:24,632 INFO [train.py:901] (0/4) Epoch 7, batch 3350, loss[loss=0.2774, simple_loss=0.3601, pruned_loss=0.09731, over 8341.00 frames. ], tot_loss[loss=0.2652, simple_loss=0.333, pruned_loss=0.09865, over 1617303.66 frames. ], batch size: 26, lr: 1.10e-02, grad_scale: 8.0 +2023-02-06 04:01:30,916 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=51857.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:01:32,338 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51859.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:01:39,525 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51869.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:01:39,585 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5619, 1.8576, 3.4356, 1.2668, 2.5707, 1.9297, 1.6626, 2.0571], + device='cuda:0'), covar=tensor([0.1509, 0.1896, 0.0515, 0.3082, 0.1120, 0.2272, 0.1493, 0.2003], + device='cuda:0'), in_proj_covar=tensor([0.0474, 0.0469, 0.0531, 0.0546, 0.0586, 0.0525, 0.0453, 0.0591], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:0') +2023-02-06 04:01:59,343 INFO [train.py:901] (0/4) Epoch 7, batch 3400, loss[loss=0.2685, simple_loss=0.3374, pruned_loss=0.09982, over 8500.00 frames. ], tot_loss[loss=0.2662, simple_loss=0.3337, pruned_loss=0.09931, over 1618529.76 frames. ], batch size: 26, lr: 1.10e-02, grad_scale: 8.0 +2023-02-06 04:02:03,608 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51905.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 04:02:20,932 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51930.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 04:02:23,259 INFO [optim.py:369] (0/4) 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 04:02:34,620 INFO [train.py:901] (0/4) Epoch 7, batch 3450, loss[loss=0.2397, simple_loss=0.2946, pruned_loss=0.09235, over 7549.00 frames. ], tot_loss[loss=0.2655, simple_loss=0.3333, pruned_loss=0.09881, over 1616956.66 frames. ], batch size: 18, lr: 1.10e-02, grad_scale: 8.0 +2023-02-06 04:02:50,954 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51972.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:02:53,155 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.86 vs. limit=2.0 +2023-02-06 04:03:09,379 INFO [train.py:901] (0/4) Epoch 7, batch 3500, loss[loss=0.2483, simple_loss=0.3169, pruned_loss=0.08983, over 7806.00 frames. ], tot_loss[loss=0.2655, simple_loss=0.3336, pruned_loss=0.0987, over 1621562.46 frames. ], batch size: 20, lr: 1.10e-02, grad_scale: 8.0 +2023-02-06 04:03:10,082 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-52000.pt +2023-02-06 04:03:22,438 WARNING [train.py:1067] (0/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-06 04:03:33,496 INFO [optim.py:369] (0/4) 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 04:03:43,726 INFO [train.py:901] (0/4) Epoch 7, batch 3550, loss[loss=0.2467, simple_loss=0.3243, pruned_loss=0.08457, over 8367.00 frames. ], tot_loss[loss=0.2648, simple_loss=0.3333, pruned_loss=0.09811, over 1618418.43 frames. ], batch size: 24, lr: 1.10e-02, grad_scale: 8.0 +2023-02-06 04:03:49,970 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52058.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:04:19,843 INFO [train.py:901] (0/4) Epoch 7, batch 3600, loss[loss=0.2183, simple_loss=0.2902, pruned_loss=0.07316, over 7798.00 frames. ], tot_loss[loss=0.2647, simple_loss=0.333, pruned_loss=0.09821, over 1620049.80 frames. ], batch size: 19, lr: 1.10e-02, grad_scale: 8.0 +2023-02-06 04:04:26,697 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=52109.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:04:30,884 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52115.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:04:37,693 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52125.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:04:43,470 INFO [optim.py:369] (0/4) 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 04:04:47,734 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52140.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:04:53,394 INFO [train.py:901] (0/4) Epoch 7, batch 3650, loss[loss=0.2692, simple_loss=0.3345, pruned_loss=0.102, over 7791.00 frames. ], tot_loss[loss=0.2656, simple_loss=0.3334, pruned_loss=0.09887, over 1619773.35 frames. ], batch size: 19, lr: 1.10e-02, grad_scale: 8.0 +2023-02-06 04:04:54,172 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52150.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:05:23,192 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 7, batch 3700, loss[loss=0.2149, simple_loss=0.2878, pruned_loss=0.07102, over 7425.00 frames. ], tot_loss[loss=0.2653, simple_loss=0.3335, pruned_loss=0.09858, over 1623522.93 frames. ], batch size: 17, lr: 1.10e-02, grad_scale: 8.0 +2023-02-06 04:05:36,882 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52211.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:05:41,634 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8221, 2.3881, 4.7497, 1.2625, 3.1641, 2.1897, 1.8303, 2.8757], + device='cuda:0'), covar=tensor([0.1451, 0.1891, 0.0540, 0.3375, 0.1234, 0.2461, 0.1553, 0.2326], + device='cuda:0'), in_proj_covar=tensor([0.0474, 0.0469, 0.0530, 0.0550, 0.0586, 0.0523, 0.0455, 0.0597], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:0') +2023-02-06 04:05:47,075 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52224.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:05:49,859 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52228.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:05:53,727 INFO [optim.py:369] (0/4) 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 04:06:04,143 INFO [train.py:901] (0/4) Epoch 7, batch 3750, loss[loss=0.2596, simple_loss=0.3216, pruned_loss=0.09879, over 7790.00 frames. ], tot_loss[loss=0.2641, simple_loss=0.3325, pruned_loss=0.09782, over 1620472.62 frames. ], batch size: 19, lr: 1.10e-02, grad_scale: 8.0 +2023-02-06 04:06:05,821 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7241, 2.3798, 4.3726, 1.2262, 2.7178, 1.9396, 1.7228, 2.7314], + device='cuda:0'), covar=tensor([0.1518, 0.1860, 0.0690, 0.3359, 0.1513, 0.2600, 0.1522, 0.2232], + device='cuda:0'), in_proj_covar=tensor([0.0473, 0.0471, 0.0532, 0.0550, 0.0585, 0.0525, 0.0456, 0.0596], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:0') +2023-02-06 04:06:07,189 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52253.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:06:12,925 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.9901, 3.2289, 2.4867, 4.2613, 1.9176, 2.2309, 2.5881, 3.1162], + device='cuda:0'), covar=tensor([0.0903, 0.0997, 0.1217, 0.0289, 0.1408, 0.1463, 0.1371, 0.1042], + device='cuda:0'), in_proj_covar=tensor([0.0254, 0.0238, 0.0280, 0.0223, 0.0241, 0.0266, 0.0274, 0.0241], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 04:06:24,074 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.27 vs. limit=5.0 +2023-02-06 04:06:38,747 INFO [train.py:901] (0/4) Epoch 7, batch 3800, loss[loss=0.3083, simple_loss=0.3614, pruned_loss=0.1276, over 8630.00 frames. ], tot_loss[loss=0.2637, simple_loss=0.3316, pruned_loss=0.09793, over 1613061.35 frames. ], batch size: 49, lr: 1.09e-02, grad_scale: 8.0 +2023-02-06 04:07:01,885 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52330.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:07:04,415 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 7, batch 3850, loss[loss=0.2194, simple_loss=0.2961, pruned_loss=0.0713, over 7930.00 frames. ], tot_loss[loss=0.2633, simple_loss=0.3311, pruned_loss=0.09778, over 1608249.08 frames. ], batch size: 20, lr: 1.09e-02, grad_scale: 8.0 +2023-02-06 04:07:25,312 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9090, 2.5230, 4.7125, 1.5047, 3.2488, 2.4878, 1.8994, 2.6648], + device='cuda:0'), covar=tensor([0.1392, 0.1755, 0.0517, 0.3182, 0.1236, 0.2244, 0.1451, 0.2373], + device='cuda:0'), in_proj_covar=tensor([0.0474, 0.0474, 0.0533, 0.0549, 0.0591, 0.0528, 0.0458, 0.0595], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:0') +2023-02-06 04:07:30,483 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-06 04:07:49,735 INFO [train.py:901] (0/4) Epoch 7, batch 3900, loss[loss=0.2068, simple_loss=0.2783, pruned_loss=0.06762, over 6832.00 frames. ], tot_loss[loss=0.2626, simple_loss=0.3302, pruned_loss=0.09751, over 1605106.12 frames. ], batch size: 15, lr: 1.09e-02, grad_scale: 8.0 +2023-02-06 04:07:51,969 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=52402.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:08:15,067 INFO [optim.py:369] (0/4) 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 04:08:25,214 INFO [train.py:901] (0/4) Epoch 7, batch 3950, loss[loss=0.3247, simple_loss=0.3771, pruned_loss=0.1361, over 8586.00 frames. ], tot_loss[loss=0.2619, simple_loss=0.3302, pruned_loss=0.09678, over 1607706.47 frames. ], batch size: 39, lr: 1.09e-02, grad_scale: 8.0 +2023-02-06 04:08:48,028 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52480.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:09:00,579 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 +2023-02-06 04:09:00,800 INFO [train.py:901] (0/4) Epoch 7, batch 4000, loss[loss=0.318, simple_loss=0.3938, pruned_loss=0.121, over 8257.00 frames. ], tot_loss[loss=0.261, simple_loss=0.3291, pruned_loss=0.09647, over 1605797.47 frames. ], batch size: 24, lr: 1.09e-02, grad_scale: 8.0 +2023-02-06 04:09:04,495 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8353, 2.4993, 2.9016, 1.0033, 2.9856, 1.6993, 1.5687, 1.6936], + device='cuda:0'), covar=tensor([0.0433, 0.0175, 0.0141, 0.0390, 0.0251, 0.0442, 0.0402, 0.0256], + device='cuda:0'), in_proj_covar=tensor([0.0352, 0.0257, 0.0215, 0.0317, 0.0258, 0.0403, 0.0311, 0.0292], + device='cuda:0'), out_proj_covar=tensor([1.1301e-04, 8.0308e-05, 6.6344e-05, 9.9245e-05, 8.1370e-05, 1.3735e-04, + 9.9871e-05, 9.1926e-05], device='cuda:0') +2023-02-06 04:09:05,153 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52505.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:09:13,165 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52517.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:09:23,936 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 +2023-02-06 04:09:24,184 INFO [optim.py:369] (0/4) 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 04:09:28,444 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.9325, 3.1883, 2.6350, 3.9120, 1.9752, 2.0750, 2.4513, 3.4003], + device='cuda:0'), covar=tensor([0.0732, 0.0990, 0.1136, 0.0296, 0.1429, 0.1447, 0.1288, 0.0685], + device='cuda:0'), in_proj_covar=tensor([0.0258, 0.0240, 0.0282, 0.0224, 0.0243, 0.0268, 0.0278, 0.0240], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 04:09:35,738 INFO [train.py:901] (0/4) Epoch 7, batch 4050, loss[loss=0.2187, simple_loss=0.2911, pruned_loss=0.0731, over 7533.00 frames. ], tot_loss[loss=0.2609, simple_loss=0.329, pruned_loss=0.09638, over 1601999.57 frames. ], batch size: 18, lr: 1.09e-02, grad_scale: 8.0 +2023-02-06 04:09:39,787 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=52555.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:09:52,443 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52573.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 04:10:10,473 INFO [train.py:901] (0/4) Epoch 7, batch 4100, loss[loss=0.2667, simple_loss=0.3295, pruned_loss=0.102, over 8041.00 frames. ], tot_loss[loss=0.2623, simple_loss=0.33, pruned_loss=0.09725, over 1604060.61 frames. ], batch size: 22, lr: 1.09e-02, grad_scale: 8.0 +2023-02-06 04:10:33,849 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 7, batch 4150, loss[loss=0.2321, simple_loss=0.3095, pruned_loss=0.07733, over 8464.00 frames. ], tot_loss[loss=0.2635, simple_loss=0.3309, pruned_loss=0.09808, over 1604164.96 frames. ], batch size: 25, lr: 1.09e-02, grad_scale: 8.0 +2023-02-06 04:10:59,828 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52670.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:11:02,369 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=52674.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:11:20,441 INFO [train.py:901] (0/4) Epoch 7, batch 4200, loss[loss=0.3327, simple_loss=0.3687, pruned_loss=0.1484, over 7549.00 frames. ], tot_loss[loss=0.2631, simple_loss=0.331, pruned_loss=0.09758, over 1611554.41 frames. ], batch size: 18, lr: 1.09e-02, grad_scale: 8.0 +2023-02-06 04:11:30,521 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-06 04:11:43,883 INFO [optim.py:369] (0/4) 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 04:11:53,178 WARNING [train.py:1067] (0/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-06 04:11:53,853 INFO [train.py:901] (0/4) Epoch 7, batch 4250, loss[loss=0.2837, simple_loss=0.3532, pruned_loss=0.1071, over 8025.00 frames. ], tot_loss[loss=0.2632, simple_loss=0.3313, pruned_loss=0.09757, over 1614219.87 frames. ], batch size: 22, lr: 1.09e-02, grad_scale: 8.0 +2023-02-06 04:12:08,145 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52770.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:12:10,305 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52773.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:12:22,411 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52789.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:12:28,478 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52798.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:12:28,937 INFO [train.py:901] (0/4) Epoch 7, batch 4300, loss[loss=0.2613, simple_loss=0.3257, pruned_loss=0.09849, over 7919.00 frames. ], tot_loss[loss=0.2609, simple_loss=0.3298, pruned_loss=0.096, over 1612764.98 frames. ], batch size: 20, lr: 1.09e-02, grad_scale: 8.0 +2023-02-06 04:12:53,669 INFO [optim.py:369] (0/4) 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 04:13:02,725 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1353, 2.4583, 1.9082, 2.9663, 1.4383, 1.5211, 1.8891, 2.3661], + device='cuda:0'), covar=tensor([0.0965, 0.0989, 0.1346, 0.0472, 0.1461, 0.1927, 0.1507, 0.1033], + device='cuda:0'), in_proj_covar=tensor([0.0254, 0.0241, 0.0281, 0.0224, 0.0237, 0.0270, 0.0278, 0.0240], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 04:13:03,877 INFO [train.py:901] (0/4) Epoch 7, batch 4350, loss[loss=0.2291, simple_loss=0.295, pruned_loss=0.08162, over 7683.00 frames. ], tot_loss[loss=0.2601, simple_loss=0.3293, pruned_loss=0.09544, over 1614280.31 frames. ], batch size: 18, lr: 1.09e-02, grad_scale: 8.0 +2023-02-06 04:13:24,436 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 7, batch 4400, loss[loss=0.2248, simple_loss=0.2919, pruned_loss=0.07884, over 7792.00 frames. ], tot_loss[loss=0.2583, simple_loss=0.3275, pruned_loss=0.09457, over 1614371.95 frames. ], batch size: 19, lr: 1.09e-02, grad_scale: 8.0 +2023-02-06 04:13:50,741 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=52917.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 04:13:56,759 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52926.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:14:02,504 INFO [optim.py:369] (0/4) 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 04:14:06,672 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-06 04:14:13,201 INFO [train.py:901] (0/4) Epoch 7, batch 4450, loss[loss=0.3643, simple_loss=0.393, pruned_loss=0.1678, over 7815.00 frames. ], tot_loss[loss=0.2593, simple_loss=0.3281, pruned_loss=0.09526, over 1612824.53 frames. ], batch size: 20, lr: 1.09e-02, grad_scale: 8.0 +2023-02-06 04:14:14,716 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52951.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:14:46,591 INFO [train.py:901] (0/4) Epoch 7, batch 4500, loss[loss=0.2102, simple_loss=0.2767, pruned_loss=0.07181, over 7543.00 frames. ], tot_loss[loss=0.2599, simple_loss=0.3285, pruned_loss=0.0957, over 1613249.14 frames. ], batch size: 18, lr: 1.09e-02, grad_scale: 8.0 +2023-02-06 04:14:59,381 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-06 04:15:10,352 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53032.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 04:15:11,481 INFO [optim.py:369] (0/4) 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 04:15:18,995 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53045.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:15:22,047 INFO [train.py:901] (0/4) Epoch 7, batch 4550, loss[loss=0.29, simple_loss=0.3534, pruned_loss=0.1133, over 8561.00 frames. ], tot_loss[loss=0.2616, simple_loss=0.3297, pruned_loss=0.09671, over 1615667.63 frames. ], batch size: 31, lr: 1.09e-02, grad_scale: 8.0 +2023-02-06 04:15:36,997 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53070.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:15:39,626 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53074.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:15:56,222 INFO [train.py:901] (0/4) Epoch 7, batch 4600, loss[loss=0.249, simple_loss=0.3065, pruned_loss=0.0958, over 7430.00 frames. ], tot_loss[loss=0.2616, simple_loss=0.3294, pruned_loss=0.09688, over 1610450.88 frames. ], batch size: 17, lr: 1.09e-02, grad_scale: 8.0 +2023-02-06 04:16:06,490 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=53114.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:16:20,973 INFO [optim.py:369] (0/4) 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 04:16:31,869 INFO [train.py:901] (0/4) Epoch 7, batch 4650, loss[loss=0.2276, simple_loss=0.2947, pruned_loss=0.08027, over 7778.00 frames. ], tot_loss[loss=0.2616, simple_loss=0.3291, pruned_loss=0.09701, over 1606166.47 frames. ], batch size: 19, lr: 1.09e-02, grad_scale: 8.0 +2023-02-06 04:16:43,018 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=2.03 vs. limit=2.0 +2023-02-06 04:17:03,262 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 +2023-02-06 04:17:06,942 INFO [train.py:901] (0/4) Epoch 7, batch 4700, loss[loss=0.2067, simple_loss=0.2838, pruned_loss=0.06481, over 7653.00 frames. ], tot_loss[loss=0.2606, simple_loss=0.3291, pruned_loss=0.09609, over 1612759.55 frames. ], batch size: 19, lr: 1.09e-02, grad_scale: 8.0 +2023-02-06 04:17:27,457 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53229.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:17:30,525 INFO [optim.py:369] (0/4) 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 04:17:41,201 INFO [train.py:901] (0/4) Epoch 7, batch 4750, loss[loss=0.317, simple_loss=0.373, pruned_loss=0.1304, over 8547.00 frames. ], tot_loss[loss=0.2622, simple_loss=0.3306, pruned_loss=0.09692, over 1612164.27 frames. ], batch size: 31, lr: 1.09e-02, grad_scale: 8.0 +2023-02-06 04:17:48,713 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53259.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:17:56,664 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-06 04:17:59,377 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-06 04:18:09,718 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53288.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 04:18:16,599 INFO [train.py:901] (0/4) Epoch 7, batch 4800, loss[loss=0.256, simple_loss=0.3344, pruned_loss=0.0888, over 8099.00 frames. ], tot_loss[loss=0.2625, simple_loss=0.3309, pruned_loss=0.09708, over 1611138.05 frames. ], batch size: 23, lr: 1.08e-02, grad_scale: 8.0 +2023-02-06 04:18:26,459 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53313.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 04:18:32,424 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53322.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:18:39,896 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 04:18:50,453 INFO [train.py:901] (0/4) Epoch 7, batch 4850, loss[loss=0.2933, simple_loss=0.3605, pruned_loss=0.113, over 8030.00 frames. ], tot_loss[loss=0.2612, simple_loss=0.3299, pruned_loss=0.09623, over 1611478.70 frames. ], batch size: 22, lr: 1.08e-02, grad_scale: 8.0 +2023-02-06 04:18:51,169 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-06 04:19:06,226 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3351, 1.5485, 1.2889, 1.9049, 0.8182, 1.1265, 1.3878, 1.4827], + device='cuda:0'), covar=tensor([0.1018, 0.0906, 0.1486, 0.0651, 0.1452, 0.1971, 0.0986, 0.0897], + device='cuda:0'), in_proj_covar=tensor([0.0253, 0.0241, 0.0278, 0.0226, 0.0241, 0.0268, 0.0280, 0.0238], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 04:19:16,474 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 7, batch 4900, loss[loss=0.2069, simple_loss=0.2741, pruned_loss=0.06991, over 7335.00 frames. ], tot_loss[loss=0.2618, simple_loss=0.3301, pruned_loss=0.09676, over 1614390.35 frames. ], batch size: 16, lr: 1.08e-02, grad_scale: 8.0 +2023-02-06 04:19:40,202 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=53418.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:19:40,289 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2462, 1.3743, 2.2726, 1.1820, 2.1250, 2.5283, 2.4643, 2.0947], + device='cuda:0'), covar=tensor([0.0939, 0.1050, 0.0506, 0.1842, 0.0545, 0.0369, 0.0612, 0.0802], + device='cuda:0'), in_proj_covar=tensor([0.0241, 0.0278, 0.0235, 0.0275, 0.0243, 0.0219, 0.0279, 0.0284], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 04:19:48,325 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.3705, 1.3222, 4.4798, 1.6838, 3.9381, 3.7586, 4.0749, 3.9346], + device='cuda:0'), covar=tensor([0.0319, 0.3710, 0.0330, 0.2968, 0.0850, 0.0633, 0.0405, 0.0498], + device='cuda:0'), in_proj_covar=tensor([0.0387, 0.0530, 0.0478, 0.0465, 0.0527, 0.0442, 0.0433, 0.0496], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 04:19:49,287 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 +2023-02-06 04:19:51,454 INFO [optim.py:369] (0/4) 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 04:19:53,023 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0103, 4.1274, 2.5536, 2.7552, 2.9124, 2.2103, 2.7291, 2.8659], + device='cuda:0'), covar=tensor([0.1284, 0.0163, 0.0753, 0.0583, 0.0496, 0.0974, 0.0746, 0.0854], + device='cuda:0'), in_proj_covar=tensor([0.0347, 0.0233, 0.0310, 0.0299, 0.0308, 0.0316, 0.0338, 0.0316], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 04:20:00,855 INFO [train.py:901] (0/4) Epoch 7, batch 4950, loss[loss=0.2755, simple_loss=0.3393, pruned_loss=0.1059, over 8511.00 frames. ], tot_loss[loss=0.264, simple_loss=0.3322, pruned_loss=0.09792, over 1618728.29 frames. ], batch size: 28, lr: 1.08e-02, grad_scale: 8.0 +2023-02-06 04:20:27,197 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53485.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:20:36,826 INFO [train.py:901] (0/4) Epoch 7, batch 5000, loss[loss=0.2482, simple_loss=0.3285, pruned_loss=0.08395, over 8311.00 frames. ], tot_loss[loss=0.2635, simple_loss=0.3319, pruned_loss=0.09759, over 1615867.13 frames. ], batch size: 25, lr: 1.08e-02, grad_scale: 8.0 +2023-02-06 04:20:45,260 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53510.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:20:48,754 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5807, 1.9795, 2.0784, 1.2003, 2.2918, 1.3631, 0.6755, 1.7220], + device='cuda:0'), covar=tensor([0.0361, 0.0155, 0.0133, 0.0254, 0.0176, 0.0427, 0.0389, 0.0151], + device='cuda:0'), in_proj_covar=tensor([0.0354, 0.0261, 0.0218, 0.0318, 0.0258, 0.0402, 0.0312, 0.0289], + device='cuda:0'), out_proj_covar=tensor([1.1305e-04, 8.1287e-05, 6.7222e-05, 9.8440e-05, 8.1323e-05, 1.3600e-04, + 9.9618e-05, 9.0591e-05], device='cuda:0') +2023-02-06 04:21:01,979 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53533.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:21:03,201 INFO [optim.py:369] (0/4) 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 04:21:12,894 INFO [train.py:901] (0/4) Epoch 7, batch 5050, loss[loss=0.2771, simple_loss=0.3451, pruned_loss=0.1046, over 8468.00 frames. ], tot_loss[loss=0.261, simple_loss=0.3303, pruned_loss=0.09589, over 1619418.55 frames. ], batch size: 25, lr: 1.08e-02, grad_scale: 8.0 +2023-02-06 04:21:31,984 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-06 04:21:43,488 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6936, 1.5737, 3.0177, 1.3156, 2.1233, 3.3064, 3.3862, 2.7825], + device='cuda:0'), covar=tensor([0.0979, 0.1280, 0.0382, 0.1988, 0.0808, 0.0290, 0.0433, 0.0654], + device='cuda:0'), in_proj_covar=tensor([0.0237, 0.0275, 0.0234, 0.0272, 0.0241, 0.0219, 0.0276, 0.0282], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 04:21:46,780 INFO [train.py:901] (0/4) Epoch 7, batch 5100, loss[loss=0.2549, simple_loss=0.3121, pruned_loss=0.0988, over 7644.00 frames. ], tot_loss[loss=0.2632, simple_loss=0.3317, pruned_loss=0.09731, over 1615363.89 frames. ], batch size: 19, lr: 1.08e-02, grad_scale: 8.0 +2023-02-06 04:21:51,221 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=53603.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:22:13,233 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 7, batch 5150, loss[loss=0.2581, simple_loss=0.3369, pruned_loss=0.08968, over 8100.00 frames. ], tot_loss[loss=0.2624, simple_loss=0.3312, pruned_loss=0.09682, over 1615376.34 frames. ], batch size: 23, lr: 1.08e-02, grad_scale: 8.0 +2023-02-06 04:22:34,950 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=53666.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:22:42,264 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 7, batch 5200, loss[loss=0.2438, simple_loss=0.3048, pruned_loss=0.09139, over 7797.00 frames. ], tot_loss[loss=0.263, simple_loss=0.3316, pruned_loss=0.09719, over 1612327.94 frames. ], batch size: 19, lr: 1.08e-02, grad_scale: 8.0 +2023-02-06 04:22:59,250 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2943, 1.2053, 1.4332, 1.0552, 0.8197, 1.1870, 1.1720, 1.1080], + device='cuda:0'), covar=tensor([0.0575, 0.1230, 0.1615, 0.1405, 0.0553, 0.1529, 0.0654, 0.0623], + device='cuda:0'), in_proj_covar=tensor([0.0116, 0.0165, 0.0204, 0.0169, 0.0114, 0.0172, 0.0126, 0.0144], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0007, 0.0008, 0.0008, 0.0005, 0.0008, 0.0006, 0.0006], + device='cuda:0') +2023-02-06 04:23:10,750 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53718.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:23:14,969 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 +2023-02-06 04:23:17,965 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 04:23:28,939 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 7, batch 5250, loss[loss=0.3249, simple_loss=0.368, pruned_loss=0.1409, over 7011.00 frames. ], tot_loss[loss=0.2633, simple_loss=0.3315, pruned_loss=0.09759, over 1610454.32 frames. ], batch size: 71, lr: 1.08e-02, grad_scale: 8.0 +2023-02-06 04:23:54,694 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53781.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:24:00,273 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53789.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:24:02,347 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 7, batch 5300, loss[loss=0.2423, simple_loss=0.3017, pruned_loss=0.09146, over 7692.00 frames. ], tot_loss[loss=0.2629, simple_loss=0.331, pruned_loss=0.09738, over 1609412.23 frames. ], batch size: 18, lr: 1.08e-02, grad_scale: 8.0 +2023-02-06 04:24:17,527 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53814.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:24:31,354 INFO [optim.py:369] (0/4) 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 04:24:39,044 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53844.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:24:42,219 INFO [train.py:901] (0/4) Epoch 7, batch 5350, loss[loss=0.2541, simple_loss=0.3157, pruned_loss=0.09625, over 7975.00 frames. ], tot_loss[loss=0.2627, simple_loss=0.3307, pruned_loss=0.09735, over 1607896.59 frames. ], batch size: 21, lr: 1.08e-02, grad_scale: 8.0 +2023-02-06 04:24:53,451 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3785, 1.4554, 1.6110, 1.3757, 0.9243, 1.7106, 0.0381, 1.0161], + device='cuda:0'), covar=tensor([0.3635, 0.1866, 0.0963, 0.2033, 0.5833, 0.0857, 0.4002, 0.2005], + device='cuda:0'), in_proj_covar=tensor([0.0145, 0.0143, 0.0085, 0.0193, 0.0231, 0.0089, 0.0147, 0.0139], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 04:25:11,784 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3820, 1.8653, 3.3734, 1.1097, 2.3022, 1.7502, 1.3915, 2.0913], + device='cuda:0'), covar=tensor([0.1660, 0.2016, 0.0623, 0.3304, 0.1477, 0.2667, 0.1725, 0.2289], + device='cuda:0'), in_proj_covar=tensor([0.0476, 0.0475, 0.0528, 0.0549, 0.0601, 0.0530, 0.0452, 0.0588], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 04:25:17,621 INFO [train.py:901] (0/4) Epoch 7, batch 5400, loss[loss=0.2421, simple_loss=0.3156, pruned_loss=0.08435, over 7805.00 frames. ], tot_loss[loss=0.2643, simple_loss=0.332, pruned_loss=0.09834, over 1608497.86 frames. ], batch size: 20, lr: 1.08e-02, grad_scale: 8.0 +2023-02-06 04:25:41,933 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 7, batch 5450, loss[loss=0.2834, simple_loss=0.3259, pruned_loss=0.1205, over 7443.00 frames. ], tot_loss[loss=0.2631, simple_loss=0.3312, pruned_loss=0.09752, over 1610534.77 frames. ], batch size: 17, lr: 1.08e-02, grad_scale: 8.0 +2023-02-06 04:26:01,442 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1506, 1.5893, 1.4614, 1.3421, 1.0777, 1.3960, 1.5468, 1.8084], + device='cuda:0'), covar=tensor([0.0523, 0.1186, 0.1759, 0.1394, 0.0583, 0.1482, 0.0705, 0.0477], + device='cuda:0'), in_proj_covar=tensor([0.0115, 0.0166, 0.0205, 0.0169, 0.0115, 0.0173, 0.0128, 0.0144], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0008, 0.0008, 0.0005, 0.0008, 0.0006, 0.0006], + device='cuda:0') +2023-02-06 04:26:08,822 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53974.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:26:18,667 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-06 04:26:26,958 INFO [train.py:901] (0/4) Epoch 7, batch 5500, loss[loss=0.2413, simple_loss=0.3229, pruned_loss=0.07981, over 8455.00 frames. ], tot_loss[loss=0.2616, simple_loss=0.3301, pruned_loss=0.0966, over 1613794.83 frames. ], batch size: 27, lr: 1.08e-02, grad_scale: 8.0 +2023-02-06 04:26:27,119 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53999.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:26:27,656 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-54000.pt +2023-02-06 04:26:52,078 INFO [optim.py:369] (0/4) 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 04:26:53,596 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=54037.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:27:01,150 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=54048.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:27:01,610 INFO [train.py:901] (0/4) Epoch 7, batch 5550, loss[loss=0.355, simple_loss=0.3872, pruned_loss=0.1614, over 7172.00 frames. ], tot_loss[loss=0.2643, simple_loss=0.3322, pruned_loss=0.09823, over 1616749.09 frames. ], batch size: 71, lr: 1.08e-02, grad_scale: 8.0 +2023-02-06 04:27:10,518 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=54062.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:27:18,025 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=54073.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:27:21,374 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.0224, 3.0929, 3.4012, 2.3417, 1.8601, 3.5731, 0.8121, 2.1669], + device='cuda:0'), covar=tensor([0.2422, 0.1422, 0.0405, 0.2079, 0.4943, 0.0580, 0.4579, 0.1888], + device='cuda:0'), in_proj_covar=tensor([0.0142, 0.0140, 0.0083, 0.0187, 0.0225, 0.0087, 0.0147, 0.0140], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 04:27:37,055 INFO [train.py:901] (0/4) Epoch 7, batch 5600, loss[loss=0.3028, simple_loss=0.3588, pruned_loss=0.1234, over 8030.00 frames. ], tot_loss[loss=0.2652, simple_loss=0.3333, pruned_loss=0.09855, over 1617175.72 frames. ], batch size: 22, lr: 1.08e-02, grad_scale: 8.0 +2023-02-06 04:27:37,951 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=54100.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:27:55,814 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=54125.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:28:02,459 INFO [optim.py:369] (0/4) 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 04:28:12,203 INFO [train.py:901] (0/4) Epoch 7, batch 5650, loss[loss=0.2563, simple_loss=0.3272, pruned_loss=0.09275, over 7800.00 frames. ], tot_loss[loss=0.2657, simple_loss=0.3334, pruned_loss=0.09894, over 1612500.33 frames. ], batch size: 20, lr: 1.08e-02, grad_scale: 8.0 +2023-02-06 04:28:25,370 WARNING [train.py:1067] (0/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-06 04:28:25,853 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.32 vs. limit=5.0 +2023-02-06 04:28:31,830 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 +2023-02-06 04:28:41,255 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0402, 2.5772, 2.6911, 1.4603, 3.1391, 1.8041, 1.5937, 1.8809], + device='cuda:0'), covar=tensor([0.0464, 0.0187, 0.0191, 0.0370, 0.0228, 0.0511, 0.0426, 0.0264], + device='cuda:0'), in_proj_covar=tensor([0.0353, 0.0261, 0.0220, 0.0325, 0.0262, 0.0412, 0.0317, 0.0300], + device='cuda:0'), out_proj_covar=tensor([1.1201e-04, 8.1238e-05, 6.7417e-05, 1.0037e-04, 8.2197e-05, 1.3914e-04, + 1.0091e-04, 9.3669e-05], device='cuda:0') +2023-02-06 04:28:47,274 INFO [train.py:901] (0/4) Epoch 7, batch 5700, loss[loss=0.2241, simple_loss=0.2899, pruned_loss=0.07916, over 7443.00 frames. ], tot_loss[loss=0.2645, simple_loss=0.3321, pruned_loss=0.09845, over 1610161.49 frames. ], batch size: 17, lr: 1.08e-02, grad_scale: 8.0 +2023-02-06 04:29:12,994 INFO [optim.py:369] (0/4) 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 04:29:22,468 INFO [train.py:901] (0/4) Epoch 7, batch 5750, loss[loss=0.2578, simple_loss=0.3396, pruned_loss=0.08798, over 8040.00 frames. ], tot_loss[loss=0.2629, simple_loss=0.3308, pruned_loss=0.09747, over 1612307.29 frames. ], batch size: 22, lr: 1.08e-02, grad_scale: 8.0 +2023-02-06 04:29:31,461 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-06 04:29:56,263 INFO [train.py:901] (0/4) Epoch 7, batch 5800, loss[loss=0.2247, simple_loss=0.3042, pruned_loss=0.07266, over 7978.00 frames. ], tot_loss[loss=0.2637, simple_loss=0.332, pruned_loss=0.09772, over 1616380.99 frames. ], batch size: 21, lr: 1.07e-02, grad_scale: 8.0 +2023-02-06 04:30:22,036 INFO [optim.py:369] (0/4) 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 04:30:32,283 INFO [train.py:901] (0/4) Epoch 7, batch 5850, loss[loss=0.2274, simple_loss=0.3133, pruned_loss=0.07078, over 8469.00 frames. ], tot_loss[loss=0.2646, simple_loss=0.3328, pruned_loss=0.09819, over 1613164.02 frames. ], batch size: 25, lr: 1.07e-02, grad_scale: 8.0 +2023-02-06 04:31:06,117 INFO [train.py:901] (0/4) Epoch 7, batch 5900, loss[loss=0.2705, simple_loss=0.3395, pruned_loss=0.1007, over 8284.00 frames. ], tot_loss[loss=0.2629, simple_loss=0.3314, pruned_loss=0.09718, over 1614973.93 frames. ], batch size: 23, lr: 1.07e-02, grad_scale: 8.0 +2023-02-06 04:31:30,649 INFO [optim.py:369] (0/4) 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 04:31:40,692 INFO [train.py:901] (0/4) Epoch 7, batch 5950, loss[loss=0.2761, simple_loss=0.3382, pruned_loss=0.107, over 8514.00 frames. ], tot_loss[loss=0.2628, simple_loss=0.3309, pruned_loss=0.09734, over 1617983.52 frames. ], batch size: 26, lr: 1.07e-02, grad_scale: 8.0 +2023-02-06 04:32:13,857 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2297, 1.7096, 1.6261, 0.8280, 1.6708, 1.2233, 0.2776, 1.5704], + device='cuda:0'), covar=tensor([0.0245, 0.0134, 0.0108, 0.0217, 0.0155, 0.0449, 0.0368, 0.0117], + device='cuda:0'), in_proj_covar=tensor([0.0351, 0.0262, 0.0223, 0.0325, 0.0263, 0.0414, 0.0322, 0.0300], + device='cuda:0'), 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:0') +2023-02-06 04:32:14,308 INFO [train.py:901] (0/4) Epoch 7, batch 6000, loss[loss=0.2628, simple_loss=0.3333, pruned_loss=0.0961, over 8247.00 frames. ], tot_loss[loss=0.2636, simple_loss=0.3318, pruned_loss=0.09773, over 1622311.25 frames. ], batch size: 24, lr: 1.07e-02, grad_scale: 8.0 +2023-02-06 04:32:14,309 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 04:32:26,546 INFO [train.py:935] (0/4) Epoch 7, validation: loss=0.2048, simple_loss=0.3036, pruned_loss=0.05298, over 944034.00 frames. +2023-02-06 04:32:26,547 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6689MB +2023-02-06 04:32:50,868 INFO [optim.py:369] (0/4) 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 04:33:00,122 INFO [train.py:901] (0/4) Epoch 7, batch 6050, loss[loss=0.3142, simple_loss=0.3639, pruned_loss=0.1323, over 8596.00 frames. ], tot_loss[loss=0.2646, simple_loss=0.3324, pruned_loss=0.09839, over 1618626.09 frames. ], batch size: 31, lr: 1.07e-02, grad_scale: 8.0 +2023-02-06 04:33:36,263 INFO [train.py:901] (0/4) Epoch 7, batch 6100, loss[loss=0.2493, simple_loss=0.3157, pruned_loss=0.09144, over 7943.00 frames. ], tot_loss[loss=0.2651, simple_loss=0.3327, pruned_loss=0.0988, over 1616463.58 frames. ], batch size: 20, lr: 1.07e-02, grad_scale: 8.0 +2023-02-06 04:33:55,843 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.1088, 1.7752, 1.6049, 1.5552, 1.2647, 1.6086, 2.2252, 1.9264], + device='cuda:0'), covar=tensor([0.0444, 0.1164, 0.1755, 0.1347, 0.0590, 0.1532, 0.0657, 0.0586], + device='cuda:0'), in_proj_covar=tensor([0.0115, 0.0165, 0.0206, 0.0167, 0.0115, 0.0171, 0.0127, 0.0145], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0008, 0.0008, 0.0005, 0.0008, 0.0006, 0.0006], + device='cuda:0') +2023-02-06 04:34:00,446 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-06 04:34:01,816 INFO [optim.py:369] (0/4) 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 04:34:11,155 INFO [train.py:901] (0/4) Epoch 7, batch 6150, loss[loss=0.3115, simple_loss=0.3783, pruned_loss=0.1224, over 8471.00 frames. ], tot_loss[loss=0.2652, simple_loss=0.3327, pruned_loss=0.09887, over 1615284.20 frames. ], batch size: 25, lr: 1.07e-02, grad_scale: 8.0 +2023-02-06 04:34:34,017 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=54682.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 04:34:38,970 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.88 vs. limit=2.0 +2023-02-06 04:34:46,674 INFO [train.py:901] (0/4) Epoch 7, batch 6200, loss[loss=0.2242, simple_loss=0.3027, pruned_loss=0.07284, over 8184.00 frames. ], tot_loss[loss=0.2655, simple_loss=0.3332, pruned_loss=0.09886, over 1618664.74 frames. ], batch size: 23, lr: 1.07e-02, grad_scale: 8.0 +2023-02-06 04:35:12,145 INFO [optim.py:369] (0/4) 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 04:35:21,776 INFO [train.py:901] (0/4) Epoch 7, batch 6250, loss[loss=0.2333, simple_loss=0.3103, pruned_loss=0.07812, over 8243.00 frames. ], tot_loss[loss=0.2643, simple_loss=0.332, pruned_loss=0.09828, over 1619392.88 frames. ], batch size: 22, lr: 1.07e-02, grad_scale: 8.0 +2023-02-06 04:35:40,874 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3064, 2.6091, 1.9326, 2.9559, 1.5629, 1.8013, 1.9745, 2.5236], + device='cuda:0'), covar=tensor([0.0700, 0.0761, 0.1121, 0.0466, 0.1148, 0.1384, 0.1204, 0.0764], + device='cuda:0'), in_proj_covar=tensor([0.0249, 0.0235, 0.0274, 0.0225, 0.0235, 0.0264, 0.0273, 0.0238], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 04:35:55,498 INFO [train.py:901] (0/4) Epoch 7, batch 6300, loss[loss=0.2748, simple_loss=0.3369, pruned_loss=0.1064, over 7932.00 frames. ], tot_loss[loss=0.2642, simple_loss=0.3317, pruned_loss=0.09833, over 1620020.39 frames. ], batch size: 20, lr: 1.07e-02, grad_scale: 8.0 +2023-02-06 04:36:22,282 INFO [optim.py:369] (0/4) 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 04:36:32,316 INFO [train.py:901] (0/4) Epoch 7, batch 6350, loss[loss=0.2865, simple_loss=0.3571, pruned_loss=0.1079, over 8240.00 frames. ], tot_loss[loss=0.2636, simple_loss=0.3314, pruned_loss=0.0979, over 1617263.92 frames. ], batch size: 24, lr: 1.07e-02, grad_scale: 8.0 +2023-02-06 04:36:53,641 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 7, batch 6400, loss[loss=0.2765, simple_loss=0.3469, pruned_loss=0.103, over 8325.00 frames. ], tot_loss[loss=0.265, simple_loss=0.3323, pruned_loss=0.09889, over 1617440.44 frames. ], batch size: 25, lr: 1.07e-02, grad_scale: 8.0 +2023-02-06 04:37:21,359 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.5727, 2.2440, 3.5986, 2.7326, 2.8230, 2.1458, 1.6921, 1.6571], + device='cuda:0'), covar=tensor([0.3017, 0.3343, 0.0802, 0.1902, 0.1911, 0.1721, 0.1540, 0.3527], + device='cuda:0'), in_proj_covar=tensor([0.0839, 0.0782, 0.0670, 0.0771, 0.0866, 0.0724, 0.0672, 0.0712], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 04:37:31,201 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 7, batch 6450, loss[loss=0.2278, simple_loss=0.3006, pruned_loss=0.07747, over 8508.00 frames. ], tot_loss[loss=0.2632, simple_loss=0.3308, pruned_loss=0.09779, over 1615438.65 frames. ], batch size: 26, lr: 1.07e-02, grad_scale: 8.0 +2023-02-06 04:37:57,771 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0688, 1.6008, 3.1480, 1.5243, 2.1899, 3.5060, 3.4226, 3.0321], + device='cuda:0'), covar=tensor([0.0873, 0.1324, 0.0416, 0.1819, 0.0885, 0.0257, 0.0425, 0.0529], + device='cuda:0'), in_proj_covar=tensor([0.0245, 0.0278, 0.0233, 0.0272, 0.0245, 0.0218, 0.0278, 0.0280], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 04:38:15,593 INFO [train.py:901] (0/4) Epoch 7, batch 6500, loss[loss=0.2674, simple_loss=0.3477, pruned_loss=0.09357, over 8556.00 frames. ], tot_loss[loss=0.2628, simple_loss=0.3304, pruned_loss=0.0976, over 1612985.99 frames. ], batch size: 31, lr: 1.07e-02, grad_scale: 8.0 +2023-02-06 04:38:33,806 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=55026.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 04:38:39,665 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 7, batch 6550, loss[loss=0.2088, simple_loss=0.285, pruned_loss=0.06633, over 7980.00 frames. ], tot_loss[loss=0.2616, simple_loss=0.3294, pruned_loss=0.0969, over 1615889.54 frames. ], batch size: 21, lr: 1.07e-02, grad_scale: 8.0 +2023-02-06 04:39:12,193 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-06 04:39:25,812 INFO [train.py:901] (0/4) Epoch 7, batch 6600, loss[loss=0.1947, simple_loss=0.2642, pruned_loss=0.06257, over 7708.00 frames. ], tot_loss[loss=0.26, simple_loss=0.328, pruned_loss=0.09601, over 1613088.41 frames. ], batch size: 18, lr: 1.07e-02, grad_scale: 8.0 +2023-02-06 04:39:32,340 WARNING [train.py:1067] (0/4) 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] (0/4) 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 04:39:53,563 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55141.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 04:39:58,774 INFO [train.py:901] (0/4) Epoch 7, batch 6650, loss[loss=0.3198, simple_loss=0.3684, pruned_loss=0.1356, over 7252.00 frames. ], tot_loss[loss=0.2625, simple_loss=0.33, pruned_loss=0.09751, over 1617350.34 frames. ], batch size: 71, lr: 1.07e-02, grad_scale: 8.0 +2023-02-06 04:40:10,773 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55166.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:40:17,203 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 +2023-02-06 04:40:17,527 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.7583, 3.6641, 3.4013, 1.6390, 3.2867, 3.3504, 3.3750, 3.0621], + device='cuda:0'), covar=tensor([0.1007, 0.0698, 0.1104, 0.4699, 0.0974, 0.1011, 0.1313, 0.0985], + device='cuda:0'), in_proj_covar=tensor([0.0411, 0.0317, 0.0348, 0.0430, 0.0338, 0.0314, 0.0325, 0.0277], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 04:40:34,123 INFO [train.py:901] (0/4) Epoch 7, batch 6700, loss[loss=0.2589, simple_loss=0.3479, pruned_loss=0.08493, over 8467.00 frames. ], tot_loss[loss=0.2612, simple_loss=0.3293, pruned_loss=0.0966, over 1616400.29 frames. ], batch size: 27, lr: 1.07e-02, grad_scale: 8.0 +2023-02-06 04:40:51,581 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=55224.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:40:55,048 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55229.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:40:58,791 INFO [optim.py:369] (0/4) 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 04:41:07,996 INFO [train.py:901] (0/4) Epoch 7, batch 6750, loss[loss=0.2434, simple_loss=0.3166, pruned_loss=0.08508, over 7975.00 frames. ], tot_loss[loss=0.2618, simple_loss=0.33, pruned_loss=0.09683, over 1618489.51 frames. ], batch size: 21, lr: 1.07e-02, grad_scale: 8.0 +2023-02-06 04:41:42,534 INFO [train.py:901] (0/4) Epoch 7, batch 6800, loss[loss=0.3269, simple_loss=0.3819, pruned_loss=0.1359, over 8677.00 frames. ], tot_loss[loss=0.2613, simple_loss=0.3296, pruned_loss=0.09651, over 1616305.02 frames. ], batch size: 34, lr: 1.07e-02, grad_scale: 16.0 +2023-02-06 04:41:47,226 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-06 04:42:08,166 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8025, 1.5887, 2.8986, 2.3620, 2.5342, 1.5580, 1.2839, 1.1939], + device='cuda:0'), covar=tensor([0.4074, 0.3962, 0.0805, 0.1766, 0.1588, 0.2437, 0.2164, 0.3396], + device='cuda:0'), in_proj_covar=tensor([0.0816, 0.0765, 0.0660, 0.0759, 0.0848, 0.0703, 0.0658, 0.0693], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 04:42:08,554 INFO [optim.py:369] (0/4) 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 04:42:11,508 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55339.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:42:18,206 INFO [train.py:901] (0/4) Epoch 7, batch 6850, loss[loss=0.2258, simple_loss=0.303, pruned_loss=0.0743, over 7967.00 frames. ], tot_loss[loss=0.2599, simple_loss=0.3287, pruned_loss=0.09554, over 1614882.30 frames. ], batch size: 21, lr: 1.06e-02, grad_scale: 16.0 +2023-02-06 04:42:34,172 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-06 04:42:50,635 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55397.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 04:42:52,430 INFO [train.py:901] (0/4) Epoch 7, batch 6900, loss[loss=0.1772, simple_loss=0.2581, pruned_loss=0.0481, over 7526.00 frames. ], tot_loss[loss=0.2597, simple_loss=0.3289, pruned_loss=0.09524, over 1610379.18 frames. ], batch size: 18, lr: 1.06e-02, grad_scale: 8.0 +2023-02-06 04:42:53,700 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 +2023-02-06 04:43:09,619 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55422.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 04:43:19,271 INFO [optim.py:369] (0/4) 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 04:43:28,912 INFO [train.py:901] (0/4) Epoch 7, batch 6950, loss[loss=0.315, simple_loss=0.3601, pruned_loss=0.135, over 8568.00 frames. ], tot_loss[loss=0.2609, simple_loss=0.3301, pruned_loss=0.0959, over 1612539.49 frames. ], batch size: 49, lr: 1.06e-02, grad_scale: 8.0 +2023-02-06 04:43:38,693 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6867, 1.6758, 3.3980, 1.3794, 2.1102, 3.7212, 3.7149, 3.1402], + device='cuda:0'), covar=tensor([0.1193, 0.1461, 0.0327, 0.2025, 0.0985, 0.0231, 0.0422, 0.0605], + device='cuda:0'), in_proj_covar=tensor([0.0242, 0.0274, 0.0232, 0.0271, 0.0243, 0.0214, 0.0278, 0.0277], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:0') +2023-02-06 04:43:46,609 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-06 04:43:59,777 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6954, 1.9006, 1.9196, 1.4777, 2.0441, 1.5974, 1.1484, 1.7574], + device='cuda:0'), covar=tensor([0.0234, 0.0140, 0.0092, 0.0187, 0.0139, 0.0289, 0.0310, 0.0111], + device='cuda:0'), in_proj_covar=tensor([0.0347, 0.0265, 0.0222, 0.0319, 0.0262, 0.0410, 0.0320, 0.0299], + device='cuda:0'), out_proj_covar=tensor([1.0918e-04, 8.1999e-05, 6.7707e-05, 9.7117e-05, 8.1776e-05, 1.3766e-04, + 1.0075e-04, 9.2945e-05], device='cuda:0') +2023-02-06 04:44:02,255 INFO [train.py:901] (0/4) Epoch 7, batch 7000, loss[loss=0.2612, simple_loss=0.3445, pruned_loss=0.08895, over 8254.00 frames. ], tot_loss[loss=0.2608, simple_loss=0.33, pruned_loss=0.0958, over 1610059.91 frames. ], batch size: 24, lr: 1.06e-02, grad_scale: 8.0 +2023-02-06 04:44:09,455 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=55510.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:44:28,250 INFO [optim.py:369] (0/4) 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 04:44:37,065 INFO [train.py:901] (0/4) Epoch 7, batch 7050, loss[loss=0.2593, simple_loss=0.3285, pruned_loss=0.095, over 8035.00 frames. ], tot_loss[loss=0.2603, simple_loss=0.3295, pruned_loss=0.09551, over 1610312.92 frames. ], batch size: 22, lr: 1.06e-02, grad_scale: 8.0 +2023-02-06 04:44:54,228 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=55573.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:44:59,848 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.9533, 1.7298, 1.7088, 1.6396, 1.3179, 1.7661, 2.3541, 1.9501], + device='cuda:0'), covar=tensor([0.0485, 0.1245, 0.1731, 0.1338, 0.0578, 0.1432, 0.0631, 0.0556], + device='cuda:0'), in_proj_covar=tensor([0.0112, 0.0166, 0.0203, 0.0167, 0.0113, 0.0171, 0.0126, 0.0142], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0008, 0.0008, 0.0005, 0.0008, 0.0006, 0.0006], + device='cuda:0') +2023-02-06 04:45:03,889 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.8991, 1.6156, 5.9990, 2.2726, 5.3856, 5.0705, 5.5744, 5.4861], + device='cuda:0'), covar=tensor([0.0350, 0.3571, 0.0239, 0.2335, 0.0766, 0.0591, 0.0344, 0.0360], + device='cuda:0'), in_proj_covar=tensor([0.0385, 0.0525, 0.0481, 0.0464, 0.0537, 0.0441, 0.0446, 0.0500], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 04:45:09,299 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55595.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:45:11,742 INFO [train.py:901] (0/4) Epoch 7, batch 7100, loss[loss=0.2803, simple_loss=0.3547, pruned_loss=0.103, over 8356.00 frames. ], tot_loss[loss=0.2621, simple_loss=0.3312, pruned_loss=0.09648, over 1612722.41 frames. ], batch size: 24, lr: 1.06e-02, grad_scale: 8.0 +2023-02-06 04:45:23,749 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.87 vs. limit=2.0 +2023-02-06 04:45:26,282 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55620.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:45:29,481 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55625.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:45:31,781 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 +2023-02-06 04:45:36,697 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 7, batch 7150, loss[loss=0.3051, simple_loss=0.3588, pruned_loss=0.1257, over 8104.00 frames. ], tot_loss[loss=0.2639, simple_loss=0.333, pruned_loss=0.09737, over 1619873.50 frames. ], batch size: 23, lr: 1.06e-02, grad_scale: 8.0 +2023-02-06 04:46:14,207 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55688.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:46:21,660 INFO [train.py:901] (0/4) Epoch 7, batch 7200, loss[loss=0.2297, simple_loss=0.2944, pruned_loss=0.08253, over 7799.00 frames. ], tot_loss[loss=0.2631, simple_loss=0.3323, pruned_loss=0.09693, over 1619972.59 frames. ], batch size: 19, lr: 1.06e-02, grad_scale: 8.0 +2023-02-06 04:46:47,134 INFO [optim.py:369] (0/4) 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 04:46:55,811 INFO [train.py:901] (0/4) Epoch 7, batch 7250, loss[loss=0.2677, simple_loss=0.3463, pruned_loss=0.0946, over 8472.00 frames. ], tot_loss[loss=0.2625, simple_loss=0.3318, pruned_loss=0.09659, over 1618591.39 frames. ], batch size: 25, lr: 1.06e-02, grad_scale: 8.0 +2023-02-06 04:47:07,765 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55766.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:47:31,009 INFO [train.py:901] (0/4) Epoch 7, batch 7300, loss[loss=0.2376, simple_loss=0.3146, pruned_loss=0.08036, over 8129.00 frames. ], tot_loss[loss=0.2628, simple_loss=0.3314, pruned_loss=0.09711, over 1615916.48 frames. ], batch size: 22, lr: 1.06e-02, grad_scale: 8.0 +2023-02-06 04:47:55,728 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 7, batch 7350, loss[loss=0.1809, simple_loss=0.2576, pruned_loss=0.05214, over 7658.00 frames. ], tot_loss[loss=0.261, simple_loss=0.3297, pruned_loss=0.09616, over 1617471.54 frames. ], batch size: 19, lr: 1.06e-02, grad_scale: 8.0 +2023-02-06 04:48:15,846 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55866.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:48:24,830 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 +2023-02-06 04:48:26,630 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55881.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:48:27,829 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-06 04:48:40,229 INFO [train.py:901] (0/4) Epoch 7, batch 7400, loss[loss=0.2546, simple_loss=0.3187, pruned_loss=0.09525, over 8130.00 frames. ], tot_loss[loss=0.2609, simple_loss=0.3293, pruned_loss=0.09629, over 1614432.88 frames. ], batch size: 22, lr: 1.06e-02, grad_scale: 8.0 +2023-02-06 04:48:45,307 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55906.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:48:50,027 WARNING [train.py:1067] (0/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-06 04:49:05,902 INFO [optim.py:369] (0/4) 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 04:49:07,533 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9533, 2.5321, 2.7297, 1.1012, 2.8663, 1.5648, 1.4826, 1.5783], + device='cuda:0'), covar=tensor([0.0439, 0.0206, 0.0169, 0.0402, 0.0220, 0.0518, 0.0488, 0.0325], + device='cuda:0'), in_proj_covar=tensor([0.0353, 0.0269, 0.0224, 0.0325, 0.0264, 0.0415, 0.0323, 0.0305], + device='cuda:0'), out_proj_covar=tensor([1.1116e-04, 8.2890e-05, 6.8041e-05, 9.8718e-05, 8.2162e-05, 1.3894e-04, + 1.0112e-04, 9.5054e-05], device='cuda:0') +2023-02-06 04:49:11,626 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55944.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:49:14,918 INFO [train.py:901] (0/4) Epoch 7, batch 7450, loss[loss=0.242, simple_loss=0.3229, pruned_loss=0.08058, over 8479.00 frames. ], tot_loss[loss=0.2608, simple_loss=0.3297, pruned_loss=0.09599, over 1615980.13 frames. ], batch size: 25, lr: 1.06e-02, grad_scale: 8.0 +2023-02-06 04:49:25,921 WARNING [train.py:1067] (0/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-06 04:49:28,827 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55969.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:49:50,668 INFO [train.py:901] (0/4) Epoch 7, batch 7500, loss[loss=0.2932, simple_loss=0.3645, pruned_loss=0.1109, over 8515.00 frames. ], tot_loss[loss=0.262, simple_loss=0.3306, pruned_loss=0.09664, over 1617889.96 frames. ], batch size: 28, lr: 1.06e-02, grad_scale: 8.0 +2023-02-06 04:49:51,378 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-56000.pt +2023-02-06 04:50:04,033 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5772, 2.2078, 4.3302, 1.2019, 2.6398, 2.0956, 1.6686, 2.2912], + device='cuda:0'), covar=tensor([0.1733, 0.2106, 0.0797, 0.3706, 0.1763, 0.2697, 0.1674, 0.2840], + device='cuda:0'), in_proj_covar=tensor([0.0476, 0.0477, 0.0533, 0.0559, 0.0598, 0.0532, 0.0457, 0.0597], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:0') +2023-02-06 04:50:17,131 INFO [optim.py:369] (0/4) 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 04:50:25,625 INFO [train.py:901] (0/4) Epoch 7, batch 7550, loss[loss=0.3336, simple_loss=0.3714, pruned_loss=0.1479, over 8585.00 frames. ], tot_loss[loss=0.2614, simple_loss=0.3298, pruned_loss=0.09644, over 1616888.58 frames. ], batch size: 31, lr: 1.06e-02, grad_scale: 8.0 +2023-02-06 04:50:31,719 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8411, 1.4347, 3.2596, 1.3042, 2.2666, 3.6079, 3.5674, 3.0112], + device='cuda:0'), covar=tensor([0.1147, 0.1570, 0.0385, 0.2197, 0.0835, 0.0261, 0.0476, 0.0735], + device='cuda:0'), in_proj_covar=tensor([0.0245, 0.0275, 0.0233, 0.0274, 0.0242, 0.0214, 0.0281, 0.0278], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 04:50:58,639 INFO [train.py:901] (0/4) Epoch 7, batch 7600, loss[loss=0.2755, simple_loss=0.344, pruned_loss=0.1035, over 8246.00 frames. ], tot_loss[loss=0.2605, simple_loss=0.3292, pruned_loss=0.09596, over 1611707.37 frames. ], batch size: 24, lr: 1.06e-02, grad_scale: 8.0 +2023-02-06 04:51:06,776 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=56110.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:51:25,297 INFO [optim.py:369] (0/4) 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 04:51:34,934 INFO [train.py:901] (0/4) Epoch 7, batch 7650, loss[loss=0.2737, simple_loss=0.3373, pruned_loss=0.105, over 7666.00 frames. ], tot_loss[loss=0.261, simple_loss=0.3296, pruned_loss=0.09626, over 1610789.41 frames. ], batch size: 19, lr: 1.06e-02, grad_scale: 8.0 +2023-02-06 04:51:44,431 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56163.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:52:08,620 INFO [train.py:901] (0/4) Epoch 7, batch 7700, loss[loss=0.2737, simple_loss=0.3444, pruned_loss=0.1015, over 8107.00 frames. ], tot_loss[loss=0.2619, simple_loss=0.33, pruned_loss=0.09693, over 1616649.18 frames. ], batch size: 23, lr: 1.06e-02, grad_scale: 8.0 +2023-02-06 04:52:11,648 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2585, 1.9164, 3.0246, 2.3541, 2.6065, 2.0401, 1.5279, 1.4044], + device='cuda:0'), covar=tensor([0.2620, 0.2897, 0.0670, 0.1684, 0.1415, 0.1433, 0.1458, 0.2909], + device='cuda:0'), in_proj_covar=tensor([0.0830, 0.0778, 0.0668, 0.0777, 0.0865, 0.0722, 0.0673, 0.0706], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 04:52:16,092 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=56210.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:52:26,756 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=56225.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:52:34,691 INFO [optim.py:369] (0/4) 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 04:52:34,720 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 7, batch 7750, loss[loss=0.2153, simple_loss=0.2951, pruned_loss=0.06774, over 7180.00 frames. ], tot_loss[loss=0.2616, simple_loss=0.3296, pruned_loss=0.09681, over 1617854.80 frames. ], batch size: 16, lr: 1.06e-02, grad_scale: 8.0 +2023-02-06 04:53:18,263 INFO [train.py:901] (0/4) Epoch 7, batch 7800, loss[loss=0.2276, simple_loss=0.3136, pruned_loss=0.07081, over 8195.00 frames. ], tot_loss[loss=0.2609, simple_loss=0.329, pruned_loss=0.09639, over 1613204.73 frames. ], batch size: 23, lr: 1.06e-02, grad_scale: 8.0 +2023-02-06 04:53:27,613 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.00 vs. limit=5.0 +2023-02-06 04:53:31,126 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.6271, 4.6234, 4.1835, 1.7719, 4.0980, 3.9982, 4.2204, 3.6257], + device='cuda:0'), covar=tensor([0.0653, 0.0417, 0.0843, 0.4540, 0.0673, 0.0741, 0.0987, 0.0808], + device='cuda:0'), in_proj_covar=tensor([0.0422, 0.0328, 0.0361, 0.0451, 0.0348, 0.0321, 0.0333, 0.0287], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 04:53:35,678 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=56325.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:53:42,619 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 7, batch 7850, loss[loss=0.2831, simple_loss=0.3462, pruned_loss=0.11, over 8246.00 frames. ], tot_loss[loss=0.2607, simple_loss=0.3286, pruned_loss=0.09637, over 1608151.59 frames. ], batch size: 22, lr: 1.06e-02, grad_scale: 8.0 +2023-02-06 04:54:24,871 INFO [train.py:901] (0/4) Epoch 7, batch 7900, loss[loss=0.2709, simple_loss=0.3331, pruned_loss=0.1043, over 7533.00 frames. ], tot_loss[loss=0.2611, simple_loss=0.329, pruned_loss=0.09661, over 1608656.30 frames. ], batch size: 18, lr: 1.05e-02, grad_scale: 8.0 +2023-02-06 04:54:49,425 INFO [optim.py:369] (0/4) 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 04:54:58,059 INFO [train.py:901] (0/4) Epoch 7, batch 7950, loss[loss=0.3065, simple_loss=0.3569, pruned_loss=0.128, over 8524.00 frames. ], tot_loss[loss=0.2632, simple_loss=0.3307, pruned_loss=0.09788, over 1611538.90 frames. ], batch size: 50, lr: 1.05e-02, grad_scale: 8.0 +2023-02-06 04:55:19,780 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4176, 2.5342, 1.7786, 2.1825, 2.0563, 1.4105, 1.8497, 1.9012], + device='cuda:0'), covar=tensor([0.1298, 0.0375, 0.0936, 0.0536, 0.0574, 0.1276, 0.0870, 0.0840], + device='cuda:0'), in_proj_covar=tensor([0.0349, 0.0236, 0.0312, 0.0300, 0.0307, 0.0317, 0.0336, 0.0315], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 04:55:19,805 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56481.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:55:31,651 INFO [train.py:901] (0/4) Epoch 7, batch 8000, loss[loss=0.3108, simple_loss=0.3619, pruned_loss=0.1298, over 6818.00 frames. ], tot_loss[loss=0.2638, simple_loss=0.3314, pruned_loss=0.09804, over 1609997.32 frames. ], batch size: 71, lr: 1.05e-02, grad_scale: 8.0 +2023-02-06 04:55:36,299 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0753, 3.9938, 2.8755, 2.9105, 3.4215, 2.3603, 3.0399, 2.9173], + device='cuda:0'), covar=tensor([0.1484, 0.0309, 0.0725, 0.0680, 0.0452, 0.1018, 0.0889, 0.0815], + device='cuda:0'), in_proj_covar=tensor([0.0350, 0.0239, 0.0312, 0.0301, 0.0308, 0.0317, 0.0337, 0.0317], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 04:55:36,304 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56506.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:55:36,340 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7351, 2.4878, 4.8434, 1.3732, 3.1554, 2.2432, 1.7763, 2.6747], + device='cuda:0'), covar=tensor([0.1602, 0.1859, 0.0528, 0.3442, 0.1486, 0.2667, 0.1718, 0.2429], + device='cuda:0'), in_proj_covar=tensor([0.0483, 0.0478, 0.0536, 0.0561, 0.0607, 0.0538, 0.0459, 0.0600], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003], + device='cuda:0') +2023-02-06 04:55:36,834 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=56507.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:55:56,107 INFO [optim.py:369] (0/4) 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 04:55:59,624 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 7, batch 8050, loss[loss=0.3478, simple_loss=0.3853, pruned_loss=0.1551, over 6640.00 frames. ], tot_loss[loss=0.2646, simple_loss=0.3309, pruned_loss=0.09917, over 1592653.91 frames. ], batch size: 71, lr: 1.05e-02, grad_scale: 8.0 +2023-02-06 04:56:27,848 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/epoch-7.pt +2023-02-06 04:56:39,418 WARNING [train.py:1067] (0/4) 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] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56581.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:56:43,094 INFO [train.py:901] (0/4) Epoch 8, batch 0, loss[loss=0.2834, simple_loss=0.3501, pruned_loss=0.1083, over 8334.00 frames. ], tot_loss[loss=0.2834, simple_loss=0.3501, pruned_loss=0.1083, over 8334.00 frames. ], batch size: 25, lr: 9.92e-03, grad_scale: 8.0 +2023-02-06 04:56:43,095 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 04:56:54,079 INFO [train.py:935] (0/4) Epoch 8, validation: loss=0.205, simple_loss=0.3028, pruned_loss=0.05355, over 944034.00 frames. +2023-02-06 04:56:54,080 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6689MB +2023-02-06 04:57:08,613 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-06 04:57:10,747 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56606.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:57:22,341 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=56622.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:57:28,990 INFO [train.py:901] (0/4) Epoch 8, batch 50, loss[loss=0.3111, simple_loss=0.3688, pruned_loss=0.1267, over 8499.00 frames. ], tot_loss[loss=0.2678, simple_loss=0.3356, pruned_loss=0.09996, over 368684.87 frames. ], batch size: 26, lr: 9.92e-03, grad_scale: 8.0 +2023-02-06 04:57:31,766 INFO [optim.py:369] (0/4) 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 04:57:43,145 WARNING [train.py:1067] (0/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-06 04:57:49,382 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-06 04:58:03,657 INFO [train.py:901] (0/4) Epoch 8, batch 100, loss[loss=0.2672, simple_loss=0.327, pruned_loss=0.1037, over 6989.00 frames. ], tot_loss[loss=0.2646, simple_loss=0.3333, pruned_loss=0.09791, over 645338.78 frames. ], batch size: 71, lr: 9.91e-03, grad_scale: 8.0 +2023-02-06 04:58:03,875 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.7368, 2.7700, 3.1586, 2.2079, 1.5522, 3.4432, 0.5588, 1.9744], + device='cuda:0'), covar=tensor([0.3063, 0.1870, 0.0774, 0.3625, 0.6855, 0.0330, 0.5097, 0.2609], + device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0147, 0.0086, 0.0195, 0.0235, 0.0088, 0.0144, 0.0143], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 04:58:05,727 WARNING [train.py:1067] (0/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-06 04:58:21,631 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.76 vs. limit=2.0 +2023-02-06 04:58:27,518 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.4376, 4.4823, 3.9231, 1.8159, 3.8878, 4.0404, 4.1080, 3.6370], + device='cuda:0'), covar=tensor([0.0944, 0.0622, 0.1157, 0.5302, 0.0914, 0.0923, 0.1230, 0.1053], + device='cuda:0'), in_proj_covar=tensor([0.0428, 0.0329, 0.0364, 0.0456, 0.0351, 0.0329, 0.0339, 0.0294], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 04:58:36,381 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3425, 1.2234, 1.4200, 1.1422, 0.8357, 1.2144, 1.1764, 0.9858], + device='cuda:0'), covar=tensor([0.0591, 0.1265, 0.1854, 0.1380, 0.0581, 0.1645, 0.0701, 0.0643], + device='cuda:0'), in_proj_covar=tensor([0.0112, 0.0163, 0.0202, 0.0165, 0.0113, 0.0170, 0.0124, 0.0141], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0008, 0.0008, 0.0005, 0.0008, 0.0006, 0.0006], + device='cuda:0') +2023-02-06 04:58:38,288 INFO [train.py:901] (0/4) Epoch 8, batch 150, loss[loss=0.2732, simple_loss=0.3367, pruned_loss=0.1048, over 8262.00 frames. ], tot_loss[loss=0.2621, simple_loss=0.3321, pruned_loss=0.09603, over 864542.10 frames. ], batch size: 24, lr: 9.91e-03, grad_scale: 8.0 +2023-02-06 04:58:38,550 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2394, 1.8661, 2.9432, 2.2707, 2.4641, 2.0372, 1.5332, 1.1451], + device='cuda:0'), covar=tensor([0.2682, 0.2838, 0.0699, 0.1680, 0.1336, 0.1537, 0.1410, 0.3110], + device='cuda:0'), in_proj_covar=tensor([0.0828, 0.0781, 0.0671, 0.0781, 0.0869, 0.0720, 0.0671, 0.0708], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 04:58:40,995 INFO [optim.py:369] (0/4) 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 04:59:13,819 INFO [train.py:901] (0/4) Epoch 8, batch 200, loss[loss=0.2362, simple_loss=0.3164, pruned_loss=0.07801, over 8342.00 frames. ], tot_loss[loss=0.2639, simple_loss=0.3335, pruned_loss=0.09714, over 1032215.25 frames. ], batch size: 25, lr: 9.90e-03, grad_scale: 8.0 +2023-02-06 04:59:41,297 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56821.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 04:59:48,621 INFO [train.py:901] (0/4) Epoch 8, batch 250, loss[loss=0.2691, simple_loss=0.3436, pruned_loss=0.09728, over 8097.00 frames. ], tot_loss[loss=0.261, simple_loss=0.3308, pruned_loss=0.09565, over 1158057.77 frames. ], batch size: 23, lr: 9.90e-03, grad_scale: 8.0 +2023-02-06 04:59:51,336 INFO [optim.py:369] (0/4) 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 04:59:56,870 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-06 05:00:06,242 WARNING [train.py:1067] (0/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-06 05:00:21,331 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0167, 2.1113, 1.9272, 2.3281, 1.6873, 1.6848, 1.9165, 2.2262], + device='cuda:0'), covar=tensor([0.0682, 0.0732, 0.0894, 0.0617, 0.1109, 0.1288, 0.0916, 0.0694], + device='cuda:0'), in_proj_covar=tensor([0.0251, 0.0235, 0.0275, 0.0223, 0.0234, 0.0266, 0.0271, 0.0237], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 05:00:21,368 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56878.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:00:23,883 INFO [train.py:901] (0/4) Epoch 8, batch 300, loss[loss=0.2697, simple_loss=0.3472, pruned_loss=0.09606, over 8470.00 frames. ], tot_loss[loss=0.2611, simple_loss=0.3309, pruned_loss=0.09566, over 1262970.46 frames. ], batch size: 29, lr: 9.89e-03, grad_scale: 8.0 +2023-02-06 05:00:25,998 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=56885.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 05:00:38,036 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56903.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:00:54,953 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56926.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:00:58,766 INFO [train.py:901] (0/4) Epoch 8, batch 350, loss[loss=0.3239, simple_loss=0.3642, pruned_loss=0.1417, over 6549.00 frames. ], tot_loss[loss=0.2605, simple_loss=0.3305, pruned_loss=0.09523, over 1341805.31 frames. ], batch size: 71, lr: 9.89e-03, grad_scale: 8.0 +2023-02-06 05:01:01,443 INFO [optim.py:369] (0/4) 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 05:01:17,025 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.8074, 5.9597, 5.0525, 2.4309, 5.1024, 5.6979, 5.3355, 5.2472], + device='cuda:0'), covar=tensor([0.0570, 0.0387, 0.0893, 0.4512, 0.0688, 0.0610, 0.1031, 0.0784], + device='cuda:0'), in_proj_covar=tensor([0.0414, 0.0324, 0.0354, 0.0442, 0.0343, 0.0324, 0.0335, 0.0287], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 05:01:33,158 INFO [train.py:901] (0/4) Epoch 8, batch 400, loss[loss=0.2322, simple_loss=0.2977, pruned_loss=0.08331, over 7657.00 frames. ], tot_loss[loss=0.2595, simple_loss=0.3296, pruned_loss=0.09468, over 1406128.42 frames. ], batch size: 19, lr: 9.89e-03, grad_scale: 8.0 +2023-02-06 05:01:43,241 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7876, 1.3909, 2.9575, 1.2316, 2.1418, 3.3679, 3.2872, 2.8274], + device='cuda:0'), covar=tensor([0.1093, 0.1461, 0.0414, 0.2020, 0.0825, 0.0246, 0.0503, 0.0625], + device='cuda:0'), in_proj_covar=tensor([0.0253, 0.0283, 0.0241, 0.0279, 0.0250, 0.0218, 0.0288, 0.0289], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 05:01:45,439 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57000.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 05:01:52,566 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57011.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:01:53,808 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57013.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:02:04,603 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.7374, 2.2870, 3.8688, 2.7805, 3.0871, 2.2858, 1.7795, 1.7055], + device='cuda:0'), covar=tensor([0.2897, 0.3338, 0.0754, 0.2076, 0.1652, 0.1610, 0.1448, 0.3593], + device='cuda:0'), in_proj_covar=tensor([0.0831, 0.0782, 0.0672, 0.0783, 0.0870, 0.0721, 0.0673, 0.0710], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 05:02:07,050 INFO [train.py:901] (0/4) Epoch 8, batch 450, loss[loss=0.2191, simple_loss=0.2984, pruned_loss=0.06991, over 7547.00 frames. ], tot_loss[loss=0.2603, simple_loss=0.3301, pruned_loss=0.09529, over 1455368.10 frames. ], batch size: 18, lr: 9.88e-03, grad_scale: 8.0 +2023-02-06 05:02:10,304 INFO [optim.py:369] (0/4) 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 05:02:18,427 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.1133, 2.2052, 1.5343, 1.7784, 1.8272, 1.2071, 1.5548, 1.6477], + device='cuda:0'), covar=tensor([0.1166, 0.0308, 0.0949, 0.0483, 0.0626, 0.1274, 0.0831, 0.0777], + device='cuda:0'), in_proj_covar=tensor([0.0346, 0.0233, 0.0316, 0.0300, 0.0309, 0.0316, 0.0339, 0.0314], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 05:02:41,864 INFO [train.py:901] (0/4) Epoch 8, batch 500, loss[loss=0.2277, simple_loss=0.307, pruned_loss=0.07419, over 8137.00 frames. ], tot_loss[loss=0.2585, simple_loss=0.3286, pruned_loss=0.09423, over 1490899.23 frames. ], batch size: 22, lr: 9.88e-03, grad_scale: 8.0 +2023-02-06 05:03:15,889 INFO [train.py:901] (0/4) Epoch 8, batch 550, loss[loss=0.2936, simple_loss=0.3419, pruned_loss=0.1227, over 8254.00 frames. ], tot_loss[loss=0.2581, simple_loss=0.328, pruned_loss=0.09414, over 1519549.55 frames. ], batch size: 24, lr: 9.87e-03, grad_scale: 8.0 +2023-02-06 05:03:18,518 INFO [optim.py:369] (0/4) 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 05:03:39,525 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=57165.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:03:50,882 INFO [train.py:901] (0/4) Epoch 8, batch 600, loss[loss=0.2639, simple_loss=0.337, pruned_loss=0.09536, over 8452.00 frames. ], tot_loss[loss=0.2583, simple_loss=0.3284, pruned_loss=0.09411, over 1541500.37 frames. ], batch size: 27, lr: 9.87e-03, grad_scale: 8.0 +2023-02-06 05:04:03,142 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-06 05:04:06,342 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57204.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:04:13,912 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 8, batch 650, loss[loss=0.2322, simple_loss=0.308, pruned_loss=0.07824, over 7808.00 frames. ], tot_loss[loss=0.2591, simple_loss=0.3292, pruned_loss=0.09455, over 1560909.45 frames. ], batch size: 20, lr: 9.86e-03, grad_scale: 8.0 +2023-02-06 05:04:28,382 INFO [optim.py:369] (0/4) 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 05:04:29,944 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6641, 2.1154, 2.0945, 1.2904, 2.3225, 1.3972, 0.7933, 1.6886], + device='cuda:0'), covar=tensor([0.0331, 0.0142, 0.0127, 0.0269, 0.0164, 0.0463, 0.0448, 0.0175], + device='cuda:0'), in_proj_covar=tensor([0.0353, 0.0271, 0.0224, 0.0330, 0.0266, 0.0418, 0.0325, 0.0302], + device='cuda:0'), out_proj_covar=tensor([1.1024e-04, 8.2685e-05, 6.7404e-05, 1.0004e-04, 8.2607e-05, 1.3904e-04, + 1.0134e-04, 9.3016e-05], device='cuda:0') +2023-02-06 05:04:41,268 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5901, 2.1675, 2.0993, 1.2377, 2.3797, 1.3393, 0.7675, 1.6328], + device='cuda:0'), covar=tensor([0.0322, 0.0154, 0.0133, 0.0288, 0.0163, 0.0488, 0.0410, 0.0176], + device='cuda:0'), in_proj_covar=tensor([0.0355, 0.0272, 0.0224, 0.0333, 0.0268, 0.0420, 0.0328, 0.0303], + device='cuda:0'), out_proj_covar=tensor([1.1101e-04, 8.3030e-05, 6.7443e-05, 1.0093e-04, 8.3005e-05, 1.3962e-04, + 1.0202e-04, 9.3388e-05], device='cuda:0') +2023-02-06 05:04:41,970 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57256.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 05:04:46,631 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57262.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:04:51,980 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57281.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 05:05:00,683 INFO [train.py:901] (0/4) Epoch 8, batch 700, loss[loss=0.2757, simple_loss=0.3429, pruned_loss=0.1043, over 8573.00 frames. ], tot_loss[loss=0.2597, simple_loss=0.3297, pruned_loss=0.09483, over 1576519.34 frames. ], batch size: 49, lr: 9.86e-03, grad_scale: 8.0 +2023-02-06 05:05:34,920 INFO [train.py:901] (0/4) Epoch 8, batch 750, loss[loss=0.2619, simple_loss=0.343, pruned_loss=0.09041, over 8361.00 frames. ], tot_loss[loss=0.2594, simple_loss=0.329, pruned_loss=0.09489, over 1582193.80 frames. ], batch size: 24, lr: 9.86e-03, grad_scale: 8.0 +2023-02-06 05:05:38,342 INFO [optim.py:369] (0/4) 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 05:05:49,687 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-06 05:05:51,140 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=57355.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:05:52,569 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=57357.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:05:57,758 WARNING [train.py:1067] (0/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-06 05:05:58,576 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3039, 1.4874, 1.3161, 1.8967, 0.7434, 1.1572, 1.2486, 1.4818], + device='cuda:0'), covar=tensor([0.0988, 0.0979, 0.1420, 0.0570, 0.1352, 0.1777, 0.0992, 0.0745], + device='cuda:0'), in_proj_covar=tensor([0.0253, 0.0235, 0.0275, 0.0217, 0.0232, 0.0267, 0.0271, 0.0237], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 05:06:09,703 INFO [train.py:901] (0/4) Epoch 8, batch 800, loss[loss=0.2941, simple_loss=0.3574, pruned_loss=0.1153, over 8585.00 frames. ], tot_loss[loss=0.2599, simple_loss=0.3292, pruned_loss=0.09528, over 1587625.49 frames. ], batch size: 49, lr: 9.85e-03, grad_scale: 16.0 +2023-02-06 05:06:10,113 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 +2023-02-06 05:06:11,937 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57385.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:06:33,598 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57416.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:06:44,181 INFO [train.py:901] (0/4) Epoch 8, batch 850, loss[loss=0.2261, simple_loss=0.2856, pruned_loss=0.08327, over 7434.00 frames. ], tot_loss[loss=0.2585, simple_loss=0.3283, pruned_loss=0.0944, over 1593899.89 frames. ], batch size: 17, lr: 9.85e-03, grad_scale: 16.0 +2023-02-06 05:06:46,897 INFO [optim.py:369] (0/4) 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 05:07:06,502 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3350, 1.4521, 2.2855, 1.1916, 1.4051, 1.6889, 1.4037, 1.4725], + device='cuda:0'), covar=tensor([0.1643, 0.1874, 0.0685, 0.3293, 0.1592, 0.2518, 0.1749, 0.1816], + device='cuda:0'), in_proj_covar=tensor([0.0472, 0.0473, 0.0526, 0.0550, 0.0593, 0.0525, 0.0452, 0.0591], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 05:07:11,119 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57470.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:07:12,465 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57472.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:07:12,498 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4808, 1.6116, 1.6802, 1.4497, 0.8661, 1.6663, 0.0700, 1.0585], + device='cuda:0'), covar=tensor([0.3200, 0.1940, 0.0703, 0.1841, 0.5990, 0.0862, 0.3781, 0.2139], + device='cuda:0'), in_proj_covar=tensor([0.0145, 0.0144, 0.0086, 0.0193, 0.0235, 0.0091, 0.0146, 0.0144], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 05:07:17,827 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 8, batch 900, loss[loss=0.2259, simple_loss=0.3071, pruned_loss=0.07236, over 8138.00 frames. ], tot_loss[loss=0.2579, simple_loss=0.3277, pruned_loss=0.09407, over 1599385.14 frames. ], batch size: 22, lr: 9.84e-03, grad_scale: 16.0 +2023-02-06 05:07:40,915 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 8, batch 950, loss[loss=0.2908, simple_loss=0.3498, pruned_loss=0.1159, over 8370.00 frames. ], tot_loss[loss=0.2565, simple_loss=0.3264, pruned_loss=0.0933, over 1600113.19 frames. ], batch size: 24, lr: 9.84e-03, grad_scale: 16.0 +2023-02-06 05:07:56,421 INFO [optim.py:369] (0/4) 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 05:07:56,671 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57536.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:08:02,727 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5349, 2.6510, 1.6833, 2.0582, 2.1621, 1.3732, 1.8064, 1.9888], + device='cuda:0'), covar=tensor([0.1107, 0.0262, 0.0945, 0.0561, 0.0525, 0.1198, 0.0767, 0.0784], + device='cuda:0'), in_proj_covar=tensor([0.0352, 0.0238, 0.0317, 0.0306, 0.0311, 0.0317, 0.0341, 0.0318], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 05:08:04,590 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=57548.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:08:08,603 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 +2023-02-06 05:08:13,023 INFO [zipformer.py:1185] (0/4) 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] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-06 05:08:14,513 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57561.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:08:29,124 INFO [train.py:901] (0/4) Epoch 8, batch 1000, loss[loss=0.2914, simple_loss=0.3437, pruned_loss=0.1196, over 6760.00 frames. ], tot_loss[loss=0.2569, simple_loss=0.3267, pruned_loss=0.09357, over 1604333.74 frames. ], batch size: 71, lr: 9.83e-03, grad_scale: 16.0 +2023-02-06 05:08:45,888 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=57606.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:08:47,905 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-06 05:09:00,599 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-06 05:09:03,900 INFO [train.py:901] (0/4) Epoch 8, batch 1050, loss[loss=0.3065, simple_loss=0.3567, pruned_loss=0.1281, over 8126.00 frames. ], tot_loss[loss=0.2586, simple_loss=0.3279, pruned_loss=0.09465, over 1612441.21 frames. ], batch size: 22, lr: 9.83e-03, grad_scale: 16.0 +2023-02-06 05:09:06,638 INFO [optim.py:369] (0/4) 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 05:09:10,045 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57641.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:09:24,430 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57663.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:09:26,492 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57666.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:09:30,460 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2663, 1.4683, 2.2287, 1.1588, 1.5197, 1.5048, 1.3223, 1.4023], + device='cuda:0'), covar=tensor([0.1598, 0.1726, 0.0707, 0.3180, 0.1409, 0.2517, 0.1655, 0.1774], + device='cuda:0'), in_proj_covar=tensor([0.0477, 0.0479, 0.0531, 0.0551, 0.0596, 0.0530, 0.0453, 0.0595], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 05:09:31,676 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 8, batch 1100, loss[loss=0.2616, simple_loss=0.3398, pruned_loss=0.09168, over 8608.00 frames. ], tot_loss[loss=0.2596, simple_loss=0.329, pruned_loss=0.09512, over 1617877.18 frames. ], batch size: 31, lr: 9.83e-03, grad_scale: 16.0 +2023-02-06 05:10:05,377 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57721.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:10:08,891 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57726.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:10:10,252 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57728.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:10:10,687 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-06 05:10:12,759 INFO [train.py:901] (0/4) Epoch 8, batch 1150, loss[loss=0.2391, simple_loss=0.314, pruned_loss=0.0821, over 8235.00 frames. ], tot_loss[loss=0.2578, simple_loss=0.3279, pruned_loss=0.09381, over 1618885.09 frames. ], batch size: 22, lr: 9.82e-03, grad_scale: 16.0 +2023-02-06 05:10:15,539 INFO [optim.py:369] (0/4) 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 05:10:26,700 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57751.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:10:26,761 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57751.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:10:28,117 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57753.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:10:28,126 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3514, 1.4844, 1.5845, 1.1757, 0.8297, 1.6434, 0.0867, 1.0128], + device='cuda:0'), covar=tensor([0.3281, 0.2068, 0.0815, 0.2400, 0.5894, 0.0792, 0.3649, 0.2122], + device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0143, 0.0084, 0.0190, 0.0232, 0.0089, 0.0143, 0.0141], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 05:10:32,786 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=57760.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:10:48,133 INFO [train.py:901] (0/4) Epoch 8, batch 1200, loss[loss=0.244, simple_loss=0.3249, pruned_loss=0.08152, over 8430.00 frames. ], tot_loss[loss=0.2573, simple_loss=0.328, pruned_loss=0.09329, over 1621256.31 frames. ], batch size: 29, lr: 9.82e-03, grad_scale: 16.0 +2023-02-06 05:11:12,640 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6029, 1.7793, 2.0783, 1.8259, 1.3356, 2.1306, 0.5439, 1.4741], + device='cuda:0'), covar=tensor([0.3384, 0.1667, 0.0551, 0.1817, 0.3899, 0.0594, 0.3832, 0.1961], + device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0143, 0.0084, 0.0189, 0.0230, 0.0089, 0.0143, 0.0140], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 05:11:17,880 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=57824.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:11:23,243 INFO [train.py:901] (0/4) Epoch 8, batch 1250, loss[loss=0.2532, simple_loss=0.3223, pruned_loss=0.09202, over 8466.00 frames. ], tot_loss[loss=0.2595, simple_loss=0.3291, pruned_loss=0.09498, over 1620110.05 frames. ], batch size: 25, lr: 9.81e-03, grad_scale: 16.0 +2023-02-06 05:11:25,902 INFO [optim.py:369] (0/4) 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 05:11:41,389 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=57857.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:11:50,898 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0127, 1.4644, 1.4844, 1.3088, 1.0653, 1.4322, 1.6297, 1.5398], + device='cuda:0'), covar=tensor([0.0462, 0.1041, 0.1459, 0.1178, 0.0531, 0.1306, 0.0593, 0.0507], + device='cuda:0'), in_proj_covar=tensor([0.0110, 0.0163, 0.0200, 0.0165, 0.0112, 0.0170, 0.0124, 0.0142], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0008, 0.0008, 0.0005, 0.0008, 0.0006, 0.0006], + device='cuda:0') +2023-02-06 05:11:53,669 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57875.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:11:58,367 INFO [train.py:901] (0/4) Epoch 8, batch 1300, loss[loss=0.2143, simple_loss=0.2993, pruned_loss=0.06462, over 8454.00 frames. ], tot_loss[loss=0.2587, simple_loss=0.3283, pruned_loss=0.09456, over 1617464.64 frames. ], batch size: 29, lr: 9.81e-03, grad_scale: 16.0 +2023-02-06 05:12:24,241 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57919.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:12:32,235 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57930.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:12:33,386 INFO [train.py:901] (0/4) Epoch 8, batch 1350, loss[loss=0.2553, simple_loss=0.3295, pruned_loss=0.09052, over 8329.00 frames. ], tot_loss[loss=0.2579, simple_loss=0.3277, pruned_loss=0.09404, over 1617564.25 frames. ], batch size: 26, lr: 9.80e-03, grad_scale: 16.0 +2023-02-06 05:12:36,119 INFO [optim.py:369] (0/4) 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 05:12:38,228 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57939.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:12:41,649 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57944.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:12:49,710 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57955.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:13:01,716 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57972.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:13:05,168 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57977.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:13:08,425 INFO [train.py:901] (0/4) Epoch 8, batch 1400, loss[loss=0.2647, simple_loss=0.3442, pruned_loss=0.09265, over 8598.00 frames. ], tot_loss[loss=0.2557, simple_loss=0.3258, pruned_loss=0.09273, over 1616753.63 frames. ], batch size: 31, lr: 9.80e-03, grad_scale: 16.0 +2023-02-06 05:13:20,620 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-58000.pt +2023-02-06 05:13:23,091 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58002.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:13:25,799 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1205, 1.7374, 2.6951, 2.1194, 2.3669, 1.9026, 1.5152, 1.0551], + device='cuda:0'), covar=tensor([0.2740, 0.2761, 0.0729, 0.1598, 0.1217, 0.1563, 0.1350, 0.2685], + device='cuda:0'), in_proj_covar=tensor([0.0835, 0.0785, 0.0675, 0.0778, 0.0867, 0.0720, 0.0665, 0.0709], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 05:13:41,318 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-06 05:13:43,354 INFO [train.py:901] (0/4) Epoch 8, batch 1450, loss[loss=0.255, simple_loss=0.3151, pruned_loss=0.09741, over 7930.00 frames. ], tot_loss[loss=0.2561, simple_loss=0.3259, pruned_loss=0.09311, over 1617091.43 frames. ], batch size: 20, lr: 9.80e-03, grad_scale: 16.0 +2023-02-06 05:13:46,043 INFO [optim.py:369] (0/4) 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 05:14:07,377 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4992, 2.0508, 3.0984, 2.4662, 2.5859, 2.1339, 1.6987, 1.2525], + device='cuda:0'), covar=tensor([0.2364, 0.2600, 0.0682, 0.1633, 0.1274, 0.1487, 0.1304, 0.2850], + device='cuda:0'), in_proj_covar=tensor([0.0832, 0.0784, 0.0672, 0.0774, 0.0863, 0.0721, 0.0663, 0.0710], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 05:14:18,668 INFO [train.py:901] (0/4) Epoch 8, batch 1500, loss[loss=0.2553, simple_loss=0.3396, pruned_loss=0.08554, over 8103.00 frames. ], tot_loss[loss=0.2566, simple_loss=0.3263, pruned_loss=0.09343, over 1618670.85 frames. ], batch size: 23, lr: 9.79e-03, grad_scale: 16.0 +2023-02-06 05:14:28,061 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=58095.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:14:52,812 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 8, batch 1550, loss[loss=0.3093, simple_loss=0.3599, pruned_loss=0.1293, over 6738.00 frames. ], tot_loss[loss=0.2562, simple_loss=0.3256, pruned_loss=0.09337, over 1615209.84 frames. ], batch size: 72, lr: 9.79e-03, grad_scale: 16.0 +2023-02-06 05:14:56,005 INFO [optim.py:369] (0/4) 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 05:15:03,106 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 +2023-02-06 05:15:10,031 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58156.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:15:12,039 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=58159.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:15:14,753 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 8, batch 1600, loss[loss=0.3141, simple_loss=0.3798, pruned_loss=0.1242, over 8111.00 frames. ], tot_loss[loss=0.2575, simple_loss=0.3267, pruned_loss=0.09413, over 1616927.83 frames. ], batch size: 23, lr: 9.78e-03, grad_scale: 16.0 +2023-02-06 05:15:36,605 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58210.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:15:54,025 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58220.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:16:00,054 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58228.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:16:02,605 INFO [train.py:901] (0/4) Epoch 8, batch 1650, loss[loss=0.2735, simple_loss=0.3445, pruned_loss=0.1012, over 8096.00 frames. ], tot_loss[loss=0.2576, simple_loss=0.3267, pruned_loss=0.0942, over 1616150.44 frames. ], batch size: 23, lr: 9.78e-03, grad_scale: 16.0 +2023-02-06 05:16:05,268 INFO [optim.py:369] (0/4) 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 05:16:16,841 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58253.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:16:20,347 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 +2023-02-06 05:16:30,961 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.1483, 1.7920, 1.8110, 1.5752, 1.1666, 1.7960, 2.0391, 2.0339], + device='cuda:0'), covar=tensor([0.0394, 0.1224, 0.1628, 0.1274, 0.0575, 0.1359, 0.0647, 0.0535], + device='cuda:0'), in_proj_covar=tensor([0.0110, 0.0162, 0.0200, 0.0165, 0.0112, 0.0170, 0.0124, 0.0142], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0008, 0.0008, 0.0005, 0.0008, 0.0006, 0.0006], + device='cuda:0') +2023-02-06 05:16:37,572 INFO [train.py:901] (0/4) Epoch 8, batch 1700, loss[loss=0.2261, simple_loss=0.2942, pruned_loss=0.07906, over 7649.00 frames. ], tot_loss[loss=0.2578, simple_loss=0.327, pruned_loss=0.09428, over 1613549.20 frames. ], batch size: 19, lr: 9.78e-03, grad_scale: 16.0 +2023-02-06 05:17:11,721 INFO [train.py:901] (0/4) Epoch 8, batch 1750, loss[loss=0.2435, simple_loss=0.3088, pruned_loss=0.08915, over 7665.00 frames. ], tot_loss[loss=0.2572, simple_loss=0.3266, pruned_loss=0.09394, over 1613883.59 frames. ], batch size: 19, lr: 9.77e-03, grad_scale: 16.0 +2023-02-06 05:17:15,045 INFO [optim.py:369] (0/4) 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 05:17:16,105 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.44 vs. limit=5.0 +2023-02-06 05:17:43,877 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3396, 2.5412, 1.6323, 1.9483, 2.0810, 1.2396, 1.7012, 1.7950], + device='cuda:0'), covar=tensor([0.1204, 0.0282, 0.0927, 0.0550, 0.0612, 0.1257, 0.0883, 0.0928], + device='cuda:0'), in_proj_covar=tensor([0.0345, 0.0231, 0.0311, 0.0298, 0.0305, 0.0315, 0.0336, 0.0318], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 05:17:45,746 INFO [train.py:901] (0/4) Epoch 8, batch 1800, loss[loss=0.272, simple_loss=0.3491, pruned_loss=0.09749, over 8479.00 frames. ], tot_loss[loss=0.2596, simple_loss=0.3284, pruned_loss=0.09542, over 1614331.45 frames. ], batch size: 25, lr: 9.77e-03, grad_scale: 16.0 +2023-02-06 05:18:21,312 INFO [train.py:901] (0/4) Epoch 8, batch 1850, loss[loss=0.1949, simple_loss=0.2728, pruned_loss=0.05852, over 7818.00 frames. ], tot_loss[loss=0.2588, simple_loss=0.3275, pruned_loss=0.09503, over 1613687.39 frames. ], batch size: 20, lr: 9.76e-03, grad_scale: 16.0 +2023-02-06 05:18:24,004 INFO [optim.py:369] (0/4) 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] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58466.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:18:46,288 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.6902, 2.3221, 4.7416, 2.8555, 4.2989, 4.1199, 4.4522, 4.3855], + device='cuda:0'), covar=tensor([0.0399, 0.2893, 0.0448, 0.2237, 0.0824, 0.0674, 0.0386, 0.0477], + device='cuda:0'), in_proj_covar=tensor([0.0405, 0.0528, 0.0500, 0.0474, 0.0539, 0.0452, 0.0453, 0.0511], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 05:18:55,854 INFO [train.py:901] (0/4) Epoch 8, batch 1900, loss[loss=0.3228, simple_loss=0.3701, pruned_loss=0.1378, over 7080.00 frames. ], tot_loss[loss=0.2572, simple_loss=0.3265, pruned_loss=0.09402, over 1612296.17 frames. ], batch size: 71, lr: 9.76e-03, grad_scale: 16.0 +2023-02-06 05:18:56,037 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0009, 2.2931, 1.7981, 2.7104, 1.4879, 1.6331, 1.9932, 2.2630], + device='cuda:0'), covar=tensor([0.0828, 0.0894, 0.1079, 0.0445, 0.1295, 0.1633, 0.1012, 0.0834], + device='cuda:0'), in_proj_covar=tensor([0.0246, 0.0229, 0.0263, 0.0213, 0.0232, 0.0264, 0.0264, 0.0232], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0004, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 05:19:01,996 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58491.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:19:10,025 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=58503.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:19:12,701 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=58506.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:19:19,277 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-06 05:19:28,969 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.3178, 1.4439, 4.4981, 1.6929, 3.9750, 3.7890, 4.0391, 3.9769], + device='cuda:0'), covar=tensor([0.0498, 0.3489, 0.0364, 0.2795, 0.0927, 0.0741, 0.0453, 0.0507], + device='cuda:0'), in_proj_covar=tensor([0.0406, 0.0529, 0.0504, 0.0476, 0.0541, 0.0453, 0.0454, 0.0512], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 05:19:30,158 INFO [train.py:901] (0/4) Epoch 8, batch 1950, loss[loss=0.2451, simple_loss=0.3307, pruned_loss=0.07974, over 8202.00 frames. ], tot_loss[loss=0.2573, simple_loss=0.3265, pruned_loss=0.09406, over 1613107.57 frames. ], batch size: 23, lr: 9.75e-03, grad_scale: 16.0 +2023-02-06 05:19:30,810 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-06 05:19:32,812 INFO [optim.py:369] (0/4) 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 05:19:50,865 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 8, batch 2000, loss[loss=0.2195, simple_loss=0.2903, pruned_loss=0.07435, over 7440.00 frames. ], tot_loss[loss=0.258, simple_loss=0.3269, pruned_loss=0.09457, over 1612449.79 frames. ], batch size: 17, lr: 9.75e-03, grad_scale: 8.0 +2023-02-06 05:20:25,668 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.19 vs. limit=5.0 +2023-02-06 05:20:29,385 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58618.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:20:31,402 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58621.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:20:39,661 INFO [train.py:901] (0/4) Epoch 8, batch 2050, loss[loss=0.2402, simple_loss=0.3102, pruned_loss=0.08504, over 7976.00 frames. ], tot_loss[loss=0.2587, simple_loss=0.3275, pruned_loss=0.09494, over 1610308.60 frames. ], batch size: 21, lr: 9.75e-03, grad_scale: 8.0 +2023-02-06 05:20:42,943 INFO [optim.py:369] (0/4) 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 05:21:13,667 INFO [train.py:901] (0/4) Epoch 8, batch 2100, loss[loss=0.2564, simple_loss=0.334, pruned_loss=0.08942, over 8612.00 frames. ], tot_loss[loss=0.2586, simple_loss=0.3277, pruned_loss=0.09471, over 1612964.11 frames. ], batch size: 31, lr: 9.74e-03, grad_scale: 8.0 +2023-02-06 05:21:25,813 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=58699.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:21:30,401 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.7087, 5.7947, 4.9523, 2.0667, 5.0587, 5.3617, 5.4560, 4.9833], + device='cuda:0'), covar=tensor([0.0695, 0.0369, 0.0912, 0.5057, 0.0698, 0.0767, 0.1000, 0.0611], + device='cuda:0'), in_proj_covar=tensor([0.0413, 0.0320, 0.0348, 0.0438, 0.0341, 0.0320, 0.0329, 0.0280], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 05:21:47,831 INFO [train.py:901] (0/4) Epoch 8, batch 2150, loss[loss=0.2474, simple_loss=0.3278, pruned_loss=0.08354, over 8470.00 frames. ], tot_loss[loss=0.2576, simple_loss=0.327, pruned_loss=0.09413, over 1610189.57 frames. ], batch size: 25, lr: 9.74e-03, grad_scale: 8.0 +2023-02-06 05:21:51,088 INFO [optim.py:369] (0/4) 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 05:22:04,053 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.48 vs. limit=5.0 +2023-02-06 05:22:23,683 INFO [train.py:901] (0/4) Epoch 8, batch 2200, loss[loss=0.2274, simple_loss=0.2998, pruned_loss=0.07752, over 7933.00 frames. ], tot_loss[loss=0.2596, simple_loss=0.3289, pruned_loss=0.09508, over 1618874.95 frames. ], batch size: 20, lr: 9.73e-03, grad_scale: 8.0 +2023-02-06 05:22:31,539 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=58793.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:22:58,809 INFO [train.py:901] (0/4) Epoch 8, batch 2250, loss[loss=0.236, simple_loss=0.3011, pruned_loss=0.08543, over 7530.00 frames. ], tot_loss[loss=0.2576, simple_loss=0.3273, pruned_loss=0.09397, over 1619333.28 frames. ], batch size: 18, lr: 9.73e-03, grad_scale: 8.0 +2023-02-06 05:23:02,320 INFO [optim.py:369] (0/4) 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 05:23:29,288 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58874.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:23:29,308 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4697, 1.9734, 3.3586, 1.2029, 2.6977, 1.8915, 1.5494, 2.1571], + device='cuda:0'), covar=tensor([0.1518, 0.1845, 0.0663, 0.3281, 0.1187, 0.2417, 0.1635, 0.2140], + device='cuda:0'), in_proj_covar=tensor([0.0478, 0.0484, 0.0533, 0.0560, 0.0605, 0.0533, 0.0457, 0.0594], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 05:23:31,287 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58877.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:23:32,129 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 +2023-02-06 05:23:34,466 INFO [train.py:901] (0/4) Epoch 8, batch 2300, loss[loss=0.2562, simple_loss=0.3279, pruned_loss=0.09227, over 8638.00 frames. ], tot_loss[loss=0.2563, simple_loss=0.3264, pruned_loss=0.09308, over 1613901.54 frames. ], batch size: 39, lr: 9.73e-03, grad_scale: 8.0 +2023-02-06 05:23:46,513 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58899.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:23:48,579 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58902.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:24:03,546 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 8, batch 2350, loss[loss=0.2761, simple_loss=0.3454, pruned_loss=0.1034, over 8497.00 frames. ], tot_loss[loss=0.2576, simple_loss=0.3273, pruned_loss=0.09391, over 1615792.70 frames. ], batch size: 26, lr: 9.72e-03, grad_scale: 8.0 +2023-02-06 05:24:12,940 INFO [optim.py:369] (0/4) 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 05:24:32,449 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1220, 2.7171, 3.1792, 1.2248, 3.3449, 1.9685, 1.4927, 1.8114], + device='cuda:0'), covar=tensor([0.0450, 0.0180, 0.0186, 0.0386, 0.0165, 0.0449, 0.0534, 0.0299], + device='cuda:0'), in_proj_covar=tensor([0.0348, 0.0270, 0.0224, 0.0328, 0.0267, 0.0416, 0.0321, 0.0303], + device='cuda:0'), out_proj_covar=tensor([1.0782e-04, 8.1729e-05, 6.7163e-05, 9.8482e-05, 8.2190e-05, 1.3724e-04, + 9.9038e-05, 9.2534e-05], device='cuda:0') +2023-02-06 05:24:44,042 INFO [train.py:901] (0/4) Epoch 8, batch 2400, loss[loss=0.2615, simple_loss=0.3246, pruned_loss=0.09917, over 7917.00 frames. ], tot_loss[loss=0.2581, simple_loss=0.3276, pruned_loss=0.09429, over 1614341.33 frames. ], batch size: 20, lr: 9.72e-03, grad_scale: 8.0 +2023-02-06 05:25:18,645 INFO [train.py:901] (0/4) Epoch 8, batch 2450, loss[loss=0.2621, simple_loss=0.3217, pruned_loss=0.1013, over 8232.00 frames. ], tot_loss[loss=0.257, simple_loss=0.3269, pruned_loss=0.09355, over 1615064.00 frames. ], batch size: 22, lr: 9.71e-03, grad_scale: 8.0 +2023-02-06 05:25:21,873 INFO [optim.py:369] (0/4) 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 05:25:26,088 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=59043.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:25:52,496 INFO [train.py:901] (0/4) Epoch 8, batch 2500, loss[loss=0.2569, simple_loss=0.3278, pruned_loss=0.09303, over 8460.00 frames. ], tot_loss[loss=0.2565, simple_loss=0.3264, pruned_loss=0.0933, over 1615469.68 frames. ], batch size: 29, lr: 9.71e-03, grad_scale: 8.0 +2023-02-06 05:26:15,602 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.1725, 2.9756, 3.3650, 2.0240, 1.7771, 3.5829, 0.6493, 2.1538], + device='cuda:0'), covar=tensor([0.2109, 0.1427, 0.0615, 0.2844, 0.5041, 0.0421, 0.4922, 0.2401], + device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0147, 0.0085, 0.0199, 0.0235, 0.0090, 0.0154, 0.0147], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 05:26:19,789 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 +2023-02-06 05:26:27,103 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2074, 1.8711, 2.7251, 2.1929, 2.3844, 1.9634, 1.5944, 1.0015], + device='cuda:0'), covar=tensor([0.2868, 0.2891, 0.0755, 0.1609, 0.1276, 0.1785, 0.1376, 0.3142], + device='cuda:0'), in_proj_covar=tensor([0.0846, 0.0794, 0.0684, 0.0786, 0.0883, 0.0737, 0.0672, 0.0720], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 05:26:27,546 INFO [train.py:901] (0/4) Epoch 8, batch 2550, loss[loss=0.2759, simple_loss=0.3407, pruned_loss=0.1055, over 7980.00 frames. ], tot_loss[loss=0.2554, simple_loss=0.3254, pruned_loss=0.09272, over 1615648.14 frames. ], batch size: 21, lr: 9.71e-03, grad_scale: 8.0 +2023-02-06 05:26:30,876 INFO [optim.py:369] (0/4) 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 05:26:30,960 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=59137.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:26:44,653 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7154, 1.4809, 2.0861, 1.8157, 1.9116, 1.4934, 1.2842, 1.1222], + device='cuda:0'), covar=tensor([0.1967, 0.2195, 0.0645, 0.1226, 0.1085, 0.1349, 0.1077, 0.2009], + device='cuda:0'), in_proj_covar=tensor([0.0843, 0.0793, 0.0683, 0.0787, 0.0881, 0.0735, 0.0671, 0.0719], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 05:26:45,980 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59158.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:27:03,041 INFO [train.py:901] (0/4) Epoch 8, batch 2600, loss[loss=0.3345, simple_loss=0.3821, pruned_loss=0.1435, over 8433.00 frames. ], tot_loss[loss=0.2548, simple_loss=0.3249, pruned_loss=0.09234, over 1616259.72 frames. ], batch size: 27, lr: 9.70e-03, grad_scale: 8.0 +2023-02-06 05:27:03,144 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59182.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:27:13,290 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0582, 2.3574, 1.6908, 2.7262, 1.4750, 1.4799, 1.9679, 2.3689], + device='cuda:0'), covar=tensor([0.0755, 0.0814, 0.1167, 0.0468, 0.1282, 0.1668, 0.1049, 0.0763], + device='cuda:0'), in_proj_covar=tensor([0.0245, 0.0227, 0.0263, 0.0214, 0.0231, 0.0264, 0.0266, 0.0236], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0004, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 05:27:38,229 INFO [train.py:901] (0/4) Epoch 8, batch 2650, loss[loss=0.3119, simple_loss=0.3708, pruned_loss=0.1265, over 8557.00 frames. ], tot_loss[loss=0.2554, simple_loss=0.3256, pruned_loss=0.09254, over 1618907.52 frames. ], batch size: 39, lr: 9.70e-03, grad_scale: 8.0 +2023-02-06 05:27:41,651 INFO [optim.py:369] (0/4) 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 05:27:51,997 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59252.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:27:59,925 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0187, 2.2962, 1.8516, 2.7022, 1.3784, 1.6833, 1.8429, 2.3766], + device='cuda:0'), covar=tensor([0.0795, 0.0803, 0.1076, 0.0475, 0.1312, 0.1454, 0.1090, 0.0739], + device='cuda:0'), in_proj_covar=tensor([0.0248, 0.0228, 0.0265, 0.0215, 0.0231, 0.0266, 0.0267, 0.0236], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0004, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 05:28:03,086 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 8, batch 2700, loss[loss=0.2853, simple_loss=0.3555, pruned_loss=0.1075, over 8479.00 frames. ], tot_loss[loss=0.2536, simple_loss=0.3241, pruned_loss=0.09157, over 1617141.61 frames. ], batch size: 25, lr: 9.69e-03, grad_scale: 8.0 +2023-02-06 05:28:46,774 INFO [train.py:901] (0/4) Epoch 8, batch 2750, loss[loss=0.2906, simple_loss=0.3467, pruned_loss=0.1173, over 7774.00 frames. ], tot_loss[loss=0.2538, simple_loss=0.3239, pruned_loss=0.0918, over 1611966.96 frames. ], batch size: 19, lr: 9.69e-03, grad_scale: 8.0 +2023-02-06 05:28:50,101 INFO [optim.py:369] (0/4) 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 05:29:20,730 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.0170, 4.0664, 3.6561, 1.7699, 3.5902, 3.5884, 3.7570, 3.3366], + device='cuda:0'), covar=tensor([0.1097, 0.0658, 0.1127, 0.5367, 0.0818, 0.0942, 0.1276, 0.0912], + device='cuda:0'), in_proj_covar=tensor([0.0426, 0.0332, 0.0360, 0.0451, 0.0352, 0.0329, 0.0342, 0.0289], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 05:29:22,662 INFO [train.py:901] (0/4) Epoch 8, batch 2800, loss[loss=0.2311, simple_loss=0.3072, pruned_loss=0.07752, over 8238.00 frames. ], tot_loss[loss=0.2549, simple_loss=0.3248, pruned_loss=0.09248, over 1608314.23 frames. ], batch size: 22, lr: 9.69e-03, grad_scale: 8.0 +2023-02-06 05:29:23,481 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59383.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:29:44,974 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59414.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:29:56,675 INFO [train.py:901] (0/4) Epoch 8, batch 2850, loss[loss=0.2127, simple_loss=0.2862, pruned_loss=0.06963, over 7647.00 frames. ], tot_loss[loss=0.2564, simple_loss=0.3259, pruned_loss=0.09342, over 1607853.30 frames. ], batch size: 19, lr: 9.68e-03, grad_scale: 8.0 +2023-02-06 05:30:00,149 INFO [optim.py:369] (0/4) 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 05:30:01,823 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59439.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:30:05,295 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0489, 0.9123, 1.0355, 1.0622, 0.7778, 1.1358, 0.0536, 0.8331], + device='cuda:0'), covar=tensor([0.2082, 0.1832, 0.0609, 0.1359, 0.3877, 0.0600, 0.3129, 0.1530], + device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0146, 0.0084, 0.0195, 0.0231, 0.0089, 0.0151, 0.0144], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 05:30:12,852 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0721, 1.2748, 1.2053, 0.3508, 1.2637, 0.9947, 0.1396, 1.1341], + device='cuda:0'), covar=tensor([0.0239, 0.0159, 0.0168, 0.0302, 0.0261, 0.0470, 0.0387, 0.0163], + device='cuda:0'), in_proj_covar=tensor([0.0353, 0.0274, 0.0225, 0.0327, 0.0266, 0.0413, 0.0323, 0.0304], + device='cuda:0'), out_proj_covar=tensor([1.0911e-04, 8.2662e-05, 6.7817e-05, 9.8048e-05, 8.1858e-05, 1.3586e-04, + 9.9259e-05, 9.2562e-05], device='cuda:0') +2023-02-06 05:30:32,532 INFO [train.py:901] (0/4) Epoch 8, batch 2900, loss[loss=0.2843, simple_loss=0.3642, pruned_loss=0.1022, over 8576.00 frames. ], tot_loss[loss=0.2555, simple_loss=0.326, pruned_loss=0.09255, over 1615466.03 frames. ], batch size: 31, lr: 9.68e-03, grad_scale: 8.0 +2023-02-06 05:30:51,212 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59508.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:30:57,864 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3012, 1.5144, 1.5105, 1.3052, 1.0440, 1.4394, 1.6274, 1.6220], + device='cuda:0'), covar=tensor([0.0495, 0.1165, 0.1733, 0.1340, 0.0570, 0.1537, 0.0702, 0.0568], + device='cuda:0'), in_proj_covar=tensor([0.0111, 0.0161, 0.0200, 0.0166, 0.0111, 0.0170, 0.0124, 0.0143], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0008, 0.0008, 0.0005, 0.0008, 0.0006, 0.0006], + device='cuda:0') +2023-02-06 05:30:59,185 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59520.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:31:03,063 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=59526.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:31:04,356 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-06 05:31:07,552 INFO [train.py:901] (0/4) Epoch 8, batch 2950, loss[loss=0.3157, simple_loss=0.3772, pruned_loss=0.1271, over 8285.00 frames. ], tot_loss[loss=0.2558, simple_loss=0.326, pruned_loss=0.09281, over 1610263.84 frames. ], batch size: 23, lr: 9.67e-03, grad_scale: 8.0 +2023-02-06 05:31:08,340 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59533.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:31:10,787 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 8, batch 3000, loss[loss=0.2317, simple_loss=0.321, pruned_loss=0.07119, over 8505.00 frames. ], tot_loss[loss=0.2559, simple_loss=0.3261, pruned_loss=0.09286, over 1612504.16 frames. ], batch size: 28, lr: 9.67e-03, grad_scale: 8.0 +2023-02-06 05:31:42,183 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 05:31:54,433 INFO [train.py:935] (0/4) Epoch 8, validation: loss=0.2021, simple_loss=0.3001, pruned_loss=0.05199, over 944034.00 frames. +2023-02-06 05:31:54,435 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6689MB +2023-02-06 05:32:25,108 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.7961, 2.3132, 3.8764, 2.6705, 3.0359, 2.3782, 1.8525, 1.5414], + device='cuda:0'), covar=tensor([0.2756, 0.3514, 0.0766, 0.2161, 0.1601, 0.1668, 0.1444, 0.3635], + device='cuda:0'), in_proj_covar=tensor([0.0843, 0.0798, 0.0681, 0.0784, 0.0883, 0.0735, 0.0670, 0.0720], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 05:32:30,901 INFO [train.py:901] (0/4) Epoch 8, batch 3050, loss[loss=0.2704, simple_loss=0.3398, pruned_loss=0.1005, over 8462.00 frames. ], tot_loss[loss=0.2554, simple_loss=0.3255, pruned_loss=0.09266, over 1612272.83 frames. ], batch size: 29, lr: 9.67e-03, grad_scale: 8.0 +2023-02-06 05:32:34,249 INFO [optim.py:369] (0/4) 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 05:32:35,800 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59639.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:32:37,136 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59641.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:32:45,610 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.2148, 4.1934, 3.8011, 1.8691, 3.7349, 3.6699, 3.7901, 3.4120], + device='cuda:0'), covar=tensor([0.0827, 0.0621, 0.1128, 0.4472, 0.0775, 0.0844, 0.1321, 0.0786], + device='cuda:0'), in_proj_covar=tensor([0.0427, 0.0336, 0.0357, 0.0452, 0.0351, 0.0331, 0.0341, 0.0293], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 05:32:52,395 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59664.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:33:05,112 INFO [train.py:901] (0/4) Epoch 8, batch 3100, loss[loss=0.3058, simple_loss=0.3752, pruned_loss=0.1182, over 8494.00 frames. ], tot_loss[loss=0.2558, simple_loss=0.3258, pruned_loss=0.09286, over 1611549.78 frames. ], batch size: 26, lr: 9.66e-03, grad_scale: 8.0 +2023-02-06 05:33:40,027 INFO [train.py:901] (0/4) Epoch 8, batch 3150, loss[loss=0.2411, simple_loss=0.3216, pruned_loss=0.08025, over 8291.00 frames. ], tot_loss[loss=0.2546, simple_loss=0.3244, pruned_loss=0.09239, over 1608810.36 frames. ], batch size: 23, lr: 9.66e-03, grad_scale: 8.0 +2023-02-06 05:33:43,231 INFO [optim.py:369] (0/4) 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 05:33:56,416 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1470, 2.6722, 3.0584, 1.3375, 3.4002, 2.1401, 1.5898, 1.7231], + device='cuda:0'), covar=tensor([0.0406, 0.0184, 0.0181, 0.0370, 0.0192, 0.0405, 0.0455, 0.0277], + device='cuda:0'), in_proj_covar=tensor([0.0355, 0.0277, 0.0226, 0.0328, 0.0264, 0.0418, 0.0323, 0.0303], + device='cuda:0'), out_proj_covar=tensor([1.0951e-04, 8.3622e-05, 6.7701e-05, 9.8110e-05, 8.1076e-05, 1.3753e-04, + 9.9177e-05, 9.2321e-05], device='cuda:0') +2023-02-06 05:34:14,623 INFO [train.py:901] (0/4) Epoch 8, batch 3200, loss[loss=0.2723, simple_loss=0.3492, pruned_loss=0.09767, over 8239.00 frames. ], tot_loss[loss=0.255, simple_loss=0.3246, pruned_loss=0.09265, over 1611996.55 frames. ], batch size: 22, lr: 9.65e-03, grad_scale: 8.0 +2023-02-06 05:34:50,980 INFO [train.py:901] (0/4) Epoch 8, batch 3250, loss[loss=0.2319, simple_loss=0.3145, pruned_loss=0.07467, over 8468.00 frames. ], tot_loss[loss=0.2553, simple_loss=0.3247, pruned_loss=0.09292, over 1608241.81 frames. ], batch size: 27, lr: 9.65e-03, grad_scale: 8.0 +2023-02-06 05:34:54,314 INFO [optim.py:369] (0/4) 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 05:35:13,093 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=59864.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:35:14,543 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59866.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:35:20,851 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.95 vs. limit=5.0 +2023-02-06 05:35:25,226 INFO [train.py:901] (0/4) Epoch 8, batch 3300, loss[loss=0.2755, simple_loss=0.3417, pruned_loss=0.1047, over 8499.00 frames. ], tot_loss[loss=0.2543, simple_loss=0.324, pruned_loss=0.09229, over 1609248.51 frames. ], batch size: 29, lr: 9.65e-03, grad_scale: 8.0 +2023-02-06 05:35:35,300 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59897.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:35:41,746 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59907.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:35:52,436 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59922.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:35:59,533 INFO [train.py:901] (0/4) Epoch 8, batch 3350, loss[loss=0.214, simple_loss=0.2868, pruned_loss=0.0706, over 7943.00 frames. ], tot_loss[loss=0.2548, simple_loss=0.3248, pruned_loss=0.09241, over 1610124.06 frames. ], batch size: 20, lr: 9.64e-03, grad_scale: 8.0 +2023-02-06 05:36:02,906 INFO [optim.py:369] (0/4) 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 05:36:09,851 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.5818, 1.9690, 3.1800, 2.3404, 2.8326, 2.2813, 1.8285, 1.2652], + device='cuda:0'), covar=tensor([0.2589, 0.3060, 0.0701, 0.1836, 0.1294, 0.1506, 0.1316, 0.3131], + device='cuda:0'), in_proj_covar=tensor([0.0843, 0.0809, 0.0685, 0.0789, 0.0890, 0.0741, 0.0679, 0.0723], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 05:36:24,470 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5968, 1.4817, 2.7419, 1.2126, 1.9800, 3.0362, 3.0444, 2.5343], + device='cuda:0'), covar=tensor([0.1027, 0.1307, 0.0427, 0.2012, 0.0794, 0.0302, 0.0491, 0.0670], + device='cuda:0'), in_proj_covar=tensor([0.0248, 0.0282, 0.0239, 0.0273, 0.0250, 0.0222, 0.0287, 0.0285], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 05:36:31,811 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59979.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:36:33,703 INFO [train.py:901] (0/4) Epoch 8, batch 3400, loss[loss=0.3127, simple_loss=0.3827, pruned_loss=0.1214, over 8336.00 frames. ], tot_loss[loss=0.2561, simple_loss=0.3258, pruned_loss=0.09315, over 1612984.36 frames. ], batch size: 26, lr: 9.64e-03, grad_scale: 8.0 +2023-02-06 05:36:46,612 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-60000.pt +2023-02-06 05:36:48,455 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1593, 1.8904, 2.8201, 2.2169, 2.5138, 2.0431, 1.5652, 1.1646], + device='cuda:0'), covar=tensor([0.3026, 0.3028, 0.0854, 0.1953, 0.1492, 0.1751, 0.1483, 0.3199], + device='cuda:0'), in_proj_covar=tensor([0.0848, 0.0815, 0.0691, 0.0793, 0.0897, 0.0744, 0.0682, 0.0727], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 05:37:09,546 INFO [train.py:901] (0/4) Epoch 8, batch 3450, loss[loss=0.3238, simple_loss=0.375, pruned_loss=0.1364, over 6697.00 frames. ], tot_loss[loss=0.2585, simple_loss=0.3278, pruned_loss=0.09459, over 1611562.35 frames. ], batch size: 71, lr: 9.63e-03, grad_scale: 8.0 +2023-02-06 05:37:12,868 INFO [optim.py:369] (0/4) 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 05:37:21,753 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([0.9970, 1.7117, 1.2982, 1.5823, 1.3494, 1.0641, 1.2126, 1.3641], + device='cuda:0'), covar=tensor([0.0909, 0.0382, 0.0966, 0.0458, 0.0693, 0.1218, 0.0736, 0.0697], + device='cuda:0'), in_proj_covar=tensor([0.0344, 0.0232, 0.0306, 0.0294, 0.0306, 0.0315, 0.0340, 0.0315], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 05:37:27,901 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60058.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:37:36,389 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60070.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:37:44,243 INFO [train.py:901] (0/4) Epoch 8, batch 3500, loss[loss=0.2328, simple_loss=0.303, pruned_loss=0.08131, over 7929.00 frames. ], tot_loss[loss=0.2554, simple_loss=0.325, pruned_loss=0.09287, over 1606075.60 frames. ], batch size: 20, lr: 9.63e-03, grad_scale: 8.0 +2023-02-06 05:37:45,046 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60083.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:37:53,896 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60096.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:38:02,916 WARNING [train.py:1067] (0/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-06 05:38:18,657 INFO [train.py:901] (0/4) Epoch 8, batch 3550, loss[loss=0.2266, simple_loss=0.2894, pruned_loss=0.08192, over 7529.00 frames. ], tot_loss[loss=0.2556, simple_loss=0.3253, pruned_loss=0.09293, over 1609189.94 frames. ], batch size: 18, lr: 9.63e-03, grad_scale: 8.0 +2023-02-06 05:38:22,098 INFO [optim.py:369] (0/4) 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 05:38:54,378 INFO [train.py:901] (0/4) Epoch 8, batch 3600, loss[loss=0.2685, simple_loss=0.3362, pruned_loss=0.1004, over 8357.00 frames. ], tot_loss[loss=0.2554, simple_loss=0.3249, pruned_loss=0.09293, over 1609731.93 frames. ], batch size: 26, lr: 9.62e-03, grad_scale: 8.0 +2023-02-06 05:39:08,515 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4171, 1.5126, 2.8883, 1.2479, 2.1302, 3.2326, 3.2210, 2.7421], + device='cuda:0'), covar=tensor([0.1254, 0.1499, 0.0490, 0.1969, 0.0976, 0.0307, 0.0513, 0.0689], + device='cuda:0'), in_proj_covar=tensor([0.0246, 0.0279, 0.0238, 0.0270, 0.0250, 0.0221, 0.0284, 0.0283], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 05:39:14,697 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) Epoch 8, batch 3650, loss[loss=0.2599, simple_loss=0.3338, pruned_loss=0.09299, over 7937.00 frames. ], tot_loss[loss=0.2549, simple_loss=0.3252, pruned_loss=0.09231, over 1617225.48 frames. ], batch size: 20, lr: 9.62e-03, grad_scale: 8.0 +2023-02-06 05:39:32,061 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60235.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:39:33,119 INFO [optim.py:369] (0/4) 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 05:39:42,736 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=60251.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:39:48,804 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60260.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:40:02,764 WARNING [train.py:1067] (0/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-06 05:40:04,773 INFO [train.py:901] (0/4) Epoch 8, batch 3700, loss[loss=0.2843, simple_loss=0.3429, pruned_loss=0.1129, over 8022.00 frames. ], tot_loss[loss=0.2558, simple_loss=0.326, pruned_loss=0.09283, over 1616402.84 frames. ], batch size: 22, lr: 9.61e-03, grad_scale: 8.0 +2023-02-06 05:40:34,558 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60325.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:40:39,100 INFO [train.py:901] (0/4) Epoch 8, batch 3750, loss[loss=0.1905, simple_loss=0.2693, pruned_loss=0.05587, over 7544.00 frames. ], tot_loss[loss=0.2547, simple_loss=0.3249, pruned_loss=0.09227, over 1613347.57 frames. ], batch size: 18, lr: 9.61e-03, grad_scale: 8.0 +2023-02-06 05:40:43,045 INFO [optim.py:369] (0/4) 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 05:41:02,727 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60366.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:41:13,776 INFO [train.py:901] (0/4) Epoch 8, batch 3800, loss[loss=0.2356, simple_loss=0.3015, pruned_loss=0.08487, over 7442.00 frames. ], tot_loss[loss=0.2547, simple_loss=0.3246, pruned_loss=0.09239, over 1610273.75 frames. ], batch size: 17, lr: 9.61e-03, grad_scale: 8.0 +2023-02-06 05:41:16,616 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4558, 1.8840, 3.1223, 2.3597, 2.7044, 2.1743, 1.6761, 1.2147], + device='cuda:0'), covar=tensor([0.2834, 0.3348, 0.0795, 0.2027, 0.1327, 0.1663, 0.1353, 0.3344], + device='cuda:0'), in_proj_covar=tensor([0.0836, 0.0811, 0.0692, 0.0786, 0.0892, 0.0741, 0.0680, 0.0730], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 05:41:21,924 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.2106, 1.8166, 5.2953, 2.2192, 4.7816, 4.4731, 4.9024, 4.7836], + device='cuda:0'), covar=tensor([0.0418, 0.3442, 0.0330, 0.2653, 0.0799, 0.0642, 0.0401, 0.0427], + device='cuda:0'), in_proj_covar=tensor([0.0414, 0.0537, 0.0513, 0.0481, 0.0549, 0.0467, 0.0461, 0.0514], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 05:41:28,151 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=60402.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:41:36,458 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=60414.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:41:45,866 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=60427.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:41:49,067 INFO [train.py:901] (0/4) Epoch 8, batch 3850, loss[loss=0.2592, simple_loss=0.329, pruned_loss=0.09475, over 8033.00 frames. ], tot_loss[loss=0.2554, simple_loss=0.3252, pruned_loss=0.09283, over 1609144.91 frames. ], batch size: 22, lr: 9.60e-03, grad_scale: 8.0 +2023-02-06 05:41:52,285 INFO [optim.py:369] (0/4) 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 05:41:54,194 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=60440.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:42:08,504 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-06 05:42:22,376 INFO [train.py:901] (0/4) Epoch 8, batch 3900, loss[loss=0.209, simple_loss=0.2804, pruned_loss=0.06881, over 8098.00 frames. ], tot_loss[loss=0.2558, simple_loss=0.3255, pruned_loss=0.09301, over 1612565.17 frames. ], batch size: 21, lr: 9.60e-03, grad_scale: 8.0 +2023-02-06 05:42:45,280 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4512, 1.8879, 2.0427, 1.1446, 2.1625, 1.4220, 0.6485, 1.5565], + device='cuda:0'), covar=tensor([0.0338, 0.0173, 0.0110, 0.0289, 0.0238, 0.0506, 0.0459, 0.0183], + device='cuda:0'), in_proj_covar=tensor([0.0354, 0.0278, 0.0227, 0.0336, 0.0269, 0.0430, 0.0330, 0.0307], + device='cuda:0'), out_proj_covar=tensor([1.0888e-04, 8.3411e-05, 6.7611e-05, 1.0025e-04, 8.2620e-05, 1.4129e-04, + 1.0130e-04, 9.3300e-05], device='cuda:0') +2023-02-06 05:42:46,589 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60517.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:42:55,283 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60529.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:42:57,129 INFO [train.py:901] (0/4) Epoch 8, batch 3950, loss[loss=0.2929, simple_loss=0.3553, pruned_loss=0.1152, over 8109.00 frames. ], tot_loss[loss=0.2555, simple_loss=0.3254, pruned_loss=0.09275, over 1616439.49 frames. ], batch size: 23, lr: 9.59e-03, grad_scale: 8.0 +2023-02-06 05:43:00,412 INFO [optim.py:369] (0/4) 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 05:43:03,866 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60542.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:43:13,341 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=60560.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:43:19,340 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.78 vs. limit=5.0 +2023-02-06 05:43:31,323 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60581.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:43:31,760 INFO [train.py:901] (0/4) Epoch 8, batch 4000, loss[loss=0.2399, simple_loss=0.3031, pruned_loss=0.08838, over 7803.00 frames. ], tot_loss[loss=0.2558, simple_loss=0.3257, pruned_loss=0.09292, over 1612689.16 frames. ], batch size: 19, lr: 9.59e-03, grad_scale: 16.0 +2023-02-06 05:43:48,250 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60606.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:43:59,775 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60622.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:44:06,259 INFO [train.py:901] (0/4) Epoch 8, batch 4050, loss[loss=0.2746, simple_loss=0.3419, pruned_loss=0.1037, over 8467.00 frames. ], tot_loss[loss=0.2555, simple_loss=0.3255, pruned_loss=0.09271, over 1612358.67 frames. ], batch size: 29, lr: 9.59e-03, grad_scale: 16.0 +2023-02-06 05:44:09,148 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7976, 1.5031, 3.0169, 1.1552, 2.0619, 3.2716, 3.3271, 2.7692], + device='cuda:0'), covar=tensor([0.1076, 0.1509, 0.0444, 0.2241, 0.0949, 0.0311, 0.0502, 0.0678], + device='cuda:0'), in_proj_covar=tensor([0.0245, 0.0279, 0.0238, 0.0271, 0.0248, 0.0219, 0.0285, 0.0281], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 05:44:09,649 INFO [optim.py:369] (0/4) 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 05:44:17,205 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60647.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:44:19,828 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8878, 1.5840, 3.3348, 1.3538, 2.2316, 3.6190, 3.6077, 3.0497], + device='cuda:0'), covar=tensor([0.0997, 0.1358, 0.0293, 0.1832, 0.0839, 0.0228, 0.0400, 0.0624], + device='cuda:0'), in_proj_covar=tensor([0.0244, 0.0278, 0.0237, 0.0271, 0.0248, 0.0218, 0.0284, 0.0281], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 05:44:36,977 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60675.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:44:41,544 INFO [train.py:901] (0/4) Epoch 8, batch 4100, loss[loss=0.2761, simple_loss=0.3434, pruned_loss=0.1044, over 8474.00 frames. ], tot_loss[loss=0.2544, simple_loss=0.3246, pruned_loss=0.09209, over 1613223.71 frames. ], batch size: 27, lr: 9.58e-03, grad_scale: 16.0 +2023-02-06 05:45:04,536 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 +2023-02-06 05:45:08,295 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.1595, 4.1977, 3.7146, 1.6766, 3.6571, 3.6151, 3.7822, 3.2667], + device='cuda:0'), covar=tensor([0.0913, 0.0719, 0.1311, 0.5077, 0.0976, 0.1123, 0.1655, 0.1039], + device='cuda:0'), in_proj_covar=tensor([0.0420, 0.0332, 0.0353, 0.0452, 0.0352, 0.0332, 0.0341, 0.0288], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 05:45:16,315 INFO [train.py:901] (0/4) Epoch 8, batch 4150, loss[loss=0.272, simple_loss=0.3385, pruned_loss=0.1027, over 7986.00 frames. ], tot_loss[loss=0.2539, simple_loss=0.324, pruned_loss=0.09189, over 1610817.95 frames. ], batch size: 21, lr: 9.58e-03, grad_scale: 16.0 +2023-02-06 05:45:19,083 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60736.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:45:19,620 INFO [optim.py:369] (0/4) 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 05:45:33,750 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 +2023-02-06 05:45:45,313 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 8, batch 4200, loss[loss=0.2078, simple_loss=0.2937, pruned_loss=0.06089, over 7970.00 frames. ], tot_loss[loss=0.2563, simple_loss=0.3264, pruned_loss=0.09316, over 1616682.19 frames. ], batch size: 21, lr: 9.57e-03, grad_scale: 8.0 +2023-02-06 05:45:54,021 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60785.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:46:02,583 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60798.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:46:02,603 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60798.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:46:07,984 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-06 05:46:10,873 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60810.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:46:11,504 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60811.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:46:20,154 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60823.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:46:25,966 INFO [train.py:901] (0/4) Epoch 8, batch 4250, loss[loss=0.2939, simple_loss=0.3657, pruned_loss=0.111, over 8559.00 frames. ], tot_loss[loss=0.2562, simple_loss=0.326, pruned_loss=0.0932, over 1613503.12 frames. ], batch size: 39, lr: 9.57e-03, grad_scale: 8.0 +2023-02-06 05:46:28,891 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60836.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:46:30,092 INFO [optim.py:369] (0/4) 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 05:46:30,813 WARNING [train.py:1067] (0/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-06 05:46:49,911 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8683, 2.1077, 1.7096, 2.5906, 1.3765, 1.6194, 1.8200, 2.2531], + device='cuda:0'), covar=tensor([0.0849, 0.0868, 0.1125, 0.0448, 0.1265, 0.1480, 0.1116, 0.0826], + device='cuda:0'), in_proj_covar=tensor([0.0248, 0.0233, 0.0269, 0.0220, 0.0228, 0.0265, 0.0268, 0.0235], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 05:47:01,071 INFO [train.py:901] (0/4) Epoch 8, batch 4300, loss[loss=0.2118, simple_loss=0.2957, pruned_loss=0.06396, over 8190.00 frames. ], tot_loss[loss=0.2574, simple_loss=0.3271, pruned_loss=0.09382, over 1615166.37 frames. ], batch size: 23, lr: 9.57e-03, grad_scale: 8.0 +2023-02-06 05:47:35,202 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 8, batch 4350, loss[loss=0.2115, simple_loss=0.2891, pruned_loss=0.06697, over 8038.00 frames. ], tot_loss[loss=0.256, simple_loss=0.3257, pruned_loss=0.09314, over 1614933.65 frames. ], batch size: 22, lr: 9.56e-03, grad_scale: 8.0 +2023-02-06 05:47:39,642 INFO [optim.py:369] (0/4) 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 05:47:51,971 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60956.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:47:54,693 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.6416, 1.2770, 4.7714, 1.8387, 4.1644, 3.9470, 4.3931, 4.1650], + device='cuda:0'), covar=tensor([0.0482, 0.4124, 0.0374, 0.3004, 0.0984, 0.0807, 0.0426, 0.0524], + device='cuda:0'), in_proj_covar=tensor([0.0419, 0.0540, 0.0521, 0.0490, 0.0548, 0.0469, 0.0467, 0.0520], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 05:47:59,291 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-06 05:48:10,325 INFO [train.py:901] (0/4) Epoch 8, batch 4400, loss[loss=0.2344, simple_loss=0.3041, pruned_loss=0.08231, over 8468.00 frames. ], tot_loss[loss=0.255, simple_loss=0.3249, pruned_loss=0.09252, over 1615071.16 frames. ], batch size: 27, lr: 9.56e-03, grad_scale: 8.0 +2023-02-06 05:48:33,374 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.1954, 1.3146, 2.3075, 1.0539, 1.7089, 1.5396, 1.2338, 1.5522], + device='cuda:0'), covar=tensor([0.2081, 0.2478, 0.0720, 0.4216, 0.1480, 0.3090, 0.2251, 0.1846], + device='cuda:0'), in_proj_covar=tensor([0.0471, 0.0487, 0.0522, 0.0559, 0.0598, 0.0542, 0.0455, 0.0593], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 05:48:42,544 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-06 05:48:44,683 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0086, 2.4571, 2.7436, 1.5023, 3.1320, 1.8234, 1.4374, 1.5875], + device='cuda:0'), covar=tensor([0.0404, 0.0201, 0.0158, 0.0346, 0.0196, 0.0471, 0.0476, 0.0271], + device='cuda:0'), in_proj_covar=tensor([0.0353, 0.0280, 0.0227, 0.0339, 0.0268, 0.0430, 0.0333, 0.0306], + device='cuda:0'), out_proj_covar=tensor([1.0817e-04, 8.3407e-05, 6.7626e-05, 1.0127e-04, 8.2175e-05, 1.4119e-04, + 1.0195e-04, 9.2573e-05], device='cuda:0') +2023-02-06 05:48:45,134 INFO [train.py:901] (0/4) Epoch 8, batch 4450, loss[loss=0.2639, simple_loss=0.3407, pruned_loss=0.09352, over 8507.00 frames. ], tot_loss[loss=0.2557, simple_loss=0.3253, pruned_loss=0.09301, over 1611045.23 frames. ], batch size: 26, lr: 9.55e-03, grad_scale: 8.0 +2023-02-06 05:48:49,128 INFO [optim.py:369] (0/4) 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 05:49:11,085 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 +2023-02-06 05:49:17,369 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=61079.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:49:17,967 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=61080.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:49:19,297 INFO [train.py:901] (0/4) Epoch 8, batch 4500, loss[loss=0.2528, simple_loss=0.3298, pruned_loss=0.08788, over 8101.00 frames. ], tot_loss[loss=0.2558, simple_loss=0.3252, pruned_loss=0.09326, over 1609516.45 frames. ], batch size: 23, lr: 9.55e-03, grad_scale: 8.0 +2023-02-06 05:49:36,537 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-06 05:49:53,484 INFO [train.py:901] (0/4) Epoch 8, batch 4550, loss[loss=0.3066, simple_loss=0.3605, pruned_loss=0.1264, over 6962.00 frames. ], tot_loss[loss=0.2574, simple_loss=0.3262, pruned_loss=0.09428, over 1606166.98 frames. ], batch size: 71, lr: 9.55e-03, grad_scale: 8.0 +2023-02-06 05:49:58,150 INFO [optim.py:369] (0/4) 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 05:50:29,329 INFO [train.py:901] (0/4) Epoch 8, batch 4600, loss[loss=0.2789, simple_loss=0.3334, pruned_loss=0.1122, over 8091.00 frames. ], tot_loss[loss=0.2572, simple_loss=0.3259, pruned_loss=0.09423, over 1603968.76 frames. ], batch size: 21, lr: 9.54e-03, grad_scale: 8.0 +2023-02-06 05:50:38,242 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=61195.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:51:04,224 INFO [train.py:901] (0/4) Epoch 8, batch 4650, loss[loss=0.2913, simple_loss=0.3677, pruned_loss=0.1074, over 8475.00 frames. ], tot_loss[loss=0.2558, simple_loss=0.3249, pruned_loss=0.09336, over 1603969.03 frames. ], batch size: 25, lr: 9.54e-03, grad_scale: 8.0 +2023-02-06 05:51:08,277 INFO [optim.py:369] (0/4) 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 05:51:38,689 INFO [train.py:901] (0/4) Epoch 8, batch 4700, loss[loss=0.2837, simple_loss=0.3571, pruned_loss=0.1052, over 8102.00 frames. ], tot_loss[loss=0.2561, simple_loss=0.3253, pruned_loss=0.09339, over 1603474.10 frames. ], batch size: 23, lr: 9.54e-03, grad_scale: 8.0 +2023-02-06 05:52:13,892 INFO [train.py:901] (0/4) Epoch 8, batch 4750, loss[loss=0.2182, simple_loss=0.3007, pruned_loss=0.06781, over 7968.00 frames. ], tot_loss[loss=0.2563, simple_loss=0.3261, pruned_loss=0.09323, over 1608351.98 frames. ], batch size: 21, lr: 9.53e-03, grad_scale: 8.0 +2023-02-06 05:52:17,856 INFO [optim.py:369] (0/4) 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 05:52:36,769 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-06 05:52:40,112 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-06 05:52:48,197 INFO [train.py:901] (0/4) Epoch 8, batch 4800, loss[loss=0.2235, simple_loss=0.2849, pruned_loss=0.08108, over 7249.00 frames. ], tot_loss[loss=0.2547, simple_loss=0.3248, pruned_loss=0.09228, over 1608984.65 frames. ], batch size: 16, lr: 9.53e-03, grad_scale: 8.0 +2023-02-06 05:52:53,983 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4574, 1.8543, 3.4667, 1.1919, 2.5247, 2.0314, 1.4841, 2.3027], + device='cuda:0'), covar=tensor([0.1573, 0.2021, 0.0625, 0.3501, 0.1397, 0.2348, 0.1746, 0.2069], + device='cuda:0'), in_proj_covar=tensor([0.0477, 0.0492, 0.0528, 0.0566, 0.0606, 0.0542, 0.0457, 0.0600], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 05:53:16,782 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 8, batch 4850, loss[loss=0.2231, simple_loss=0.2934, pruned_loss=0.07642, over 7639.00 frames. ], tot_loss[loss=0.2547, simple_loss=0.3245, pruned_loss=0.09243, over 1608965.60 frames. ], batch size: 19, lr: 9.52e-03, grad_scale: 8.0 +2023-02-06 05:53:26,658 INFO [optim.py:369] (0/4) 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 05:53:28,679 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-06 05:53:36,184 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=61451.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:53:53,267 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=61476.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:53:56,861 INFO [train.py:901] (0/4) Epoch 8, batch 4900, loss[loss=0.2503, simple_loss=0.3257, pruned_loss=0.08744, over 8495.00 frames. ], tot_loss[loss=0.2543, simple_loss=0.3243, pruned_loss=0.09219, over 1606326.61 frames. ], batch size: 28, lr: 9.52e-03, grad_scale: 8.0 +2023-02-06 05:54:31,292 INFO [train.py:901] (0/4) Epoch 8, batch 4950, loss[loss=0.2773, simple_loss=0.3486, pruned_loss=0.103, over 8505.00 frames. ], tot_loss[loss=0.2545, simple_loss=0.3245, pruned_loss=0.09224, over 1607358.49 frames. ], batch size: 26, lr: 9.52e-03, grad_scale: 8.0 +2023-02-06 05:54:35,324 INFO [optim.py:369] (0/4) 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 05:54:35,524 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=61538.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:54:53,018 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 +2023-02-06 05:55:07,162 INFO [train.py:901] (0/4) Epoch 8, batch 5000, loss[loss=0.211, simple_loss=0.3018, pruned_loss=0.06012, over 8365.00 frames. ], tot_loss[loss=0.2537, simple_loss=0.3238, pruned_loss=0.09184, over 1605638.91 frames. ], batch size: 24, lr: 9.51e-03, grad_scale: 8.0 +2023-02-06 05:55:15,979 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=2.09 vs. limit=2.0 +2023-02-06 05:55:25,674 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=2.05 vs. limit=2.0 +2023-02-06 05:55:42,154 INFO [train.py:901] (0/4) Epoch 8, batch 5050, loss[loss=0.265, simple_loss=0.3236, pruned_loss=0.1032, over 7534.00 frames. ], tot_loss[loss=0.2531, simple_loss=0.3229, pruned_loss=0.09161, over 1604782.74 frames. ], batch size: 18, lr: 9.51e-03, grad_scale: 8.0 +2023-02-06 05:55:46,797 INFO [optim.py:369] (0/4) 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 05:56:07,737 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-06 05:56:17,402 INFO [train.py:901] (0/4) Epoch 8, batch 5100, loss[loss=0.2011, simple_loss=0.2805, pruned_loss=0.06089, over 8237.00 frames. ], tot_loss[loss=0.2537, simple_loss=0.3233, pruned_loss=0.09206, over 1606200.97 frames. ], batch size: 22, lr: 9.50e-03, grad_scale: 8.0 +2023-02-06 05:56:19,699 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4108, 1.6277, 2.8211, 1.1389, 2.1580, 1.7238, 1.5087, 1.8258], + device='cuda:0'), covar=tensor([0.1647, 0.2190, 0.0650, 0.3712, 0.1314, 0.2764, 0.1767, 0.1920], + device='cuda:0'), in_proj_covar=tensor([0.0478, 0.0491, 0.0536, 0.0568, 0.0609, 0.0540, 0.0458, 0.0601], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 05:56:52,943 INFO [train.py:901] (0/4) Epoch 8, batch 5150, loss[loss=0.2973, simple_loss=0.3667, pruned_loss=0.114, over 8514.00 frames. ], tot_loss[loss=0.2544, simple_loss=0.3244, pruned_loss=0.09216, over 1611096.45 frames. ], batch size: 26, lr: 9.50e-03, grad_scale: 8.0 +2023-02-06 05:56:57,121 INFO [optim.py:369] (0/4) 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 05:57:05,484 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=61750.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:57:13,682 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=61761.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:57:27,544 INFO [train.py:901] (0/4) Epoch 8, batch 5200, loss[loss=0.2514, simple_loss=0.3291, pruned_loss=0.08683, over 8473.00 frames. ], tot_loss[loss=0.2562, simple_loss=0.326, pruned_loss=0.09319, over 1615730.57 frames. ], batch size: 27, lr: 9.50e-03, grad_scale: 8.0 +2023-02-06 05:57:35,703 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=61794.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:57:53,066 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=61819.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:58:02,415 INFO [train.py:901] (0/4) Epoch 8, batch 5250, loss[loss=0.1854, simple_loss=0.2645, pruned_loss=0.05313, over 7269.00 frames. ], tot_loss[loss=0.2547, simple_loss=0.3245, pruned_loss=0.0925, over 1614778.84 frames. ], batch size: 16, lr: 9.49e-03, grad_scale: 8.0 +2023-02-06 05:58:06,461 INFO [optim.py:369] (0/4) 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 05:58:07,808 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 8, batch 5300, loss[loss=0.2238, simple_loss=0.3115, pruned_loss=0.06802, over 8496.00 frames. ], tot_loss[loss=0.2538, simple_loss=0.3234, pruned_loss=0.09204, over 1610493.90 frames. ], batch size: 26, lr: 9.49e-03, grad_scale: 8.0 +2023-02-06 05:59:12,099 INFO [train.py:901] (0/4) Epoch 8, batch 5350, loss[loss=0.2264, simple_loss=0.3096, pruned_loss=0.07162, over 8471.00 frames. ], tot_loss[loss=0.2537, simple_loss=0.3237, pruned_loss=0.09184, over 1611643.26 frames. ], batch size: 25, lr: 9.49e-03, grad_scale: 8.0 +2023-02-06 05:59:12,250 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=61932.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 05:59:16,966 INFO [optim.py:369] (0/4) 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 05:59:47,806 INFO [train.py:901] (0/4) Epoch 8, batch 5400, loss[loss=0.2553, simple_loss=0.3378, pruned_loss=0.08634, over 8472.00 frames. ], tot_loss[loss=0.254, simple_loss=0.3239, pruned_loss=0.09201, over 1612506.60 frames. ], batch size: 27, lr: 9.48e-03, grad_scale: 8.0 +2023-02-06 06:00:01,018 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-62000.pt +2023-02-06 06:00:23,971 INFO [train.py:901] (0/4) Epoch 8, batch 5450, loss[loss=0.2534, simple_loss=0.3319, pruned_loss=0.08739, over 8131.00 frames. ], tot_loss[loss=0.2538, simple_loss=0.3244, pruned_loss=0.09157, over 1615855.46 frames. ], batch size: 22, lr: 9.48e-03, grad_scale: 8.0 +2023-02-06 06:00:28,693 INFO [optim.py:369] (0/4) 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 06:01:00,417 INFO [train.py:901] (0/4) Epoch 8, batch 5500, loss[loss=0.2566, simple_loss=0.3335, pruned_loss=0.0898, over 8718.00 frames. ], tot_loss[loss=0.2544, simple_loss=0.325, pruned_loss=0.09186, over 1619020.85 frames. ], batch size: 34, lr: 9.47e-03, grad_scale: 8.0 +2023-02-06 06:01:01,800 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-06 06:01:08,451 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=62094.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:01:15,827 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=62105.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:01:34,831 INFO [train.py:901] (0/4) Epoch 8, batch 5550, loss[loss=0.2294, simple_loss=0.298, pruned_loss=0.08041, over 7791.00 frames. ], tot_loss[loss=0.2542, simple_loss=0.325, pruned_loss=0.09165, over 1618581.97 frames. ], batch size: 19, lr: 9.47e-03, grad_scale: 8.0 +2023-02-06 06:01:38,602 INFO [optim.py:369] (0/4) 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 06:01:40,098 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62140.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:01:41,543 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0164, 1.2211, 1.2105, 0.4185, 1.2400, 0.9963, 0.0686, 1.0116], + device='cuda:0'), covar=tensor([0.0252, 0.0177, 0.0164, 0.0318, 0.0206, 0.0520, 0.0438, 0.0191], + device='cuda:0'), in_proj_covar=tensor([0.0347, 0.0275, 0.0225, 0.0335, 0.0264, 0.0422, 0.0324, 0.0304], + device='cuda:0'), out_proj_covar=tensor([1.0585e-04, 8.1793e-05, 6.6734e-05, 1.0025e-04, 8.0082e-05, 1.3780e-04, + 9.9031e-05, 9.1440e-05], device='cuda:0') +2023-02-06 06:02:09,237 INFO [train.py:901] (0/4) Epoch 8, batch 5600, loss[loss=0.3296, simple_loss=0.373, pruned_loss=0.1431, over 7242.00 frames. ], tot_loss[loss=0.2554, simple_loss=0.3259, pruned_loss=0.09248, over 1615327.14 frames. ], batch size: 72, lr: 9.47e-03, grad_scale: 8.0 +2023-02-06 06:02:19,258 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.0831, 1.2146, 4.1867, 1.4718, 3.6363, 3.4709, 3.7587, 3.6020], + device='cuda:0'), covar=tensor([0.0488, 0.4139, 0.0474, 0.3190, 0.1109, 0.0833, 0.0549, 0.0668], + device='cuda:0'), in_proj_covar=tensor([0.0431, 0.0540, 0.0522, 0.0496, 0.0559, 0.0470, 0.0473, 0.0525], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 06:02:27,944 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62209.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:02:35,377 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62220.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:02:42,271 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9145, 1.6065, 3.2526, 1.3974, 2.2017, 3.5703, 3.5005, 2.9533], + device='cuda:0'), covar=tensor([0.1020, 0.1369, 0.0368, 0.1901, 0.0920, 0.0257, 0.0463, 0.0728], + device='cuda:0'), in_proj_covar=tensor([0.0250, 0.0279, 0.0241, 0.0271, 0.0253, 0.0222, 0.0289, 0.0283], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 06:02:44,113 INFO [train.py:901] (0/4) Epoch 8, batch 5650, loss[loss=0.2363, simple_loss=0.3118, pruned_loss=0.08035, over 8243.00 frames. ], tot_loss[loss=0.2562, simple_loss=0.3261, pruned_loss=0.09316, over 1614680.22 frames. ], batch size: 22, lr: 9.46e-03, grad_scale: 8.0 +2023-02-06 06:02:48,206 INFO [optim.py:369] (0/4) 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 06:03:03,304 WARNING [train.py:1067] (0/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-06 06:03:14,924 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=62276.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:03:16,527 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4577, 1.9005, 3.0228, 2.4532, 2.6578, 2.1273, 1.6117, 1.3886], + device='cuda:0'), covar=tensor([0.2651, 0.3161, 0.0833, 0.1876, 0.1420, 0.1675, 0.1478, 0.3314], + device='cuda:0'), in_proj_covar=tensor([0.0849, 0.0805, 0.0679, 0.0792, 0.0893, 0.0740, 0.0675, 0.0730], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 06:03:19,746 INFO [train.py:901] (0/4) Epoch 8, batch 5700, loss[loss=0.2063, simple_loss=0.2789, pruned_loss=0.06679, over 7807.00 frames. ], tot_loss[loss=0.2553, simple_loss=0.3253, pruned_loss=0.09268, over 1615031.45 frames. ], batch size: 20, lr: 9.46e-03, grad_scale: 8.0 +2023-02-06 06:03:26,007 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62291.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:03:53,764 INFO [train.py:901] (0/4) Epoch 8, batch 5750, loss[loss=0.2369, simple_loss=0.3028, pruned_loss=0.08548, over 7976.00 frames. ], tot_loss[loss=0.2542, simple_loss=0.324, pruned_loss=0.09225, over 1614503.73 frames. ], batch size: 21, lr: 9.45e-03, grad_scale: 8.0 +2023-02-06 06:03:58,445 INFO [optim.py:369] (0/4) 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] (0/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-06 06:04:28,273 INFO [train.py:901] (0/4) Epoch 8, batch 5800, loss[loss=0.2556, simple_loss=0.3427, pruned_loss=0.08427, over 8198.00 frames. ], tot_loss[loss=0.2538, simple_loss=0.3236, pruned_loss=0.09204, over 1610729.88 frames. ], batch size: 23, lr: 9.45e-03, grad_scale: 8.0 +2023-02-06 06:04:35,226 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62391.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:04:48,144 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62409.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:04:52,278 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1168, 2.4859, 1.9038, 2.8378, 1.2843, 1.5609, 1.8506, 2.2581], + device='cuda:0'), covar=tensor([0.0963, 0.0899, 0.1291, 0.0420, 0.1488, 0.1842, 0.1330, 0.1057], + device='cuda:0'), in_proj_covar=tensor([0.0248, 0.0225, 0.0264, 0.0214, 0.0228, 0.0261, 0.0263, 0.0231], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0004, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 06:05:04,103 INFO [train.py:901] (0/4) Epoch 8, batch 5850, loss[loss=0.3017, simple_loss=0.3537, pruned_loss=0.1249, over 7003.00 frames. ], tot_loss[loss=0.2538, simple_loss=0.3236, pruned_loss=0.09198, over 1607964.55 frames. ], batch size: 71, lr: 9.45e-03, grad_scale: 8.0 +2023-02-06 06:05:04,209 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.7824, 5.8866, 5.0048, 2.5108, 5.1534, 5.6172, 5.4292, 5.1120], + device='cuda:0'), covar=tensor([0.0861, 0.0626, 0.1096, 0.4407, 0.0747, 0.0578, 0.1479, 0.0536], + device='cuda:0'), in_proj_covar=tensor([0.0435, 0.0344, 0.0363, 0.0462, 0.0361, 0.0341, 0.0350, 0.0298], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 06:05:08,211 INFO [optim.py:369] (0/4) 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 06:05:17,217 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 +2023-02-06 06:05:27,159 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62465.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:05:34,603 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 8, batch 5900, loss[loss=0.2374, simple_loss=0.3342, pruned_loss=0.07032, over 8341.00 frames. ], tot_loss[loss=0.2531, simple_loss=0.3236, pruned_loss=0.09125, over 1610004.03 frames. ], batch size: 26, lr: 9.44e-03, grad_scale: 8.0 +2023-02-06 06:05:39,866 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=62484.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:05:44,013 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62490.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:05:51,355 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62501.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:05:54,094 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0008, 2.3878, 2.7072, 1.3268, 2.7402, 1.8286, 1.6292, 1.7420], + device='cuda:0'), covar=tensor([0.0410, 0.0208, 0.0138, 0.0349, 0.0224, 0.0401, 0.0434, 0.0237], + device='cuda:0'), in_proj_covar=tensor([0.0353, 0.0283, 0.0230, 0.0341, 0.0273, 0.0432, 0.0333, 0.0310], + device='cuda:0'), out_proj_covar=tensor([1.0759e-04, 8.3933e-05, 6.7697e-05, 1.0158e-04, 8.2743e-05, 1.4112e-04, + 1.0179e-04, 9.2918e-05], device='cuda:0') +2023-02-06 06:06:13,419 INFO [train.py:901] (0/4) Epoch 8, batch 5950, loss[loss=0.3018, simple_loss=0.3601, pruned_loss=0.1218, over 7240.00 frames. ], tot_loss[loss=0.2539, simple_loss=0.3242, pruned_loss=0.09178, over 1609514.72 frames. ], batch size: 72, lr: 9.44e-03, grad_scale: 8.0 +2023-02-06 06:06:15,676 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.0848, 2.9957, 3.2018, 1.9342, 1.6585, 3.4044, 0.5464, 2.1765], + device='cuda:0'), covar=tensor([0.2002, 0.1881, 0.0544, 0.3466, 0.5612, 0.0351, 0.5704, 0.2652], + device='cuda:0'), in_proj_covar=tensor([0.0152, 0.0154, 0.0089, 0.0202, 0.0243, 0.0092, 0.0162, 0.0155], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0003, 0.0002], + device='cuda:0') +2023-02-06 06:06:17,412 INFO [optim.py:369] (0/4) 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 06:06:34,787 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 8, batch 6000, loss[loss=0.2345, simple_loss=0.2888, pruned_loss=0.09013, over 7262.00 frames. ], tot_loss[loss=0.2538, simple_loss=0.3241, pruned_loss=0.09174, over 1608096.94 frames. ], batch size: 16, lr: 9.44e-03, grad_scale: 8.0 +2023-02-06 06:06:47,824 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 06:07:00,017 INFO [train.py:935] (0/4) Epoch 8, validation: loss=0.1996, simple_loss=0.2985, pruned_loss=0.05037, over 944034.00 frames. +2023-02-06 06:07:00,018 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6689MB +2023-02-06 06:07:12,292 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62599.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:07:14,954 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62603.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:07:33,808 INFO [train.py:901] (0/4) Epoch 8, batch 6050, loss[loss=0.2136, simple_loss=0.2881, pruned_loss=0.0696, over 7804.00 frames. ], tot_loss[loss=0.2561, simple_loss=0.326, pruned_loss=0.09307, over 1610109.38 frames. ], batch size: 20, lr: 9.43e-03, grad_scale: 8.0 +2023-02-06 06:07:35,914 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=62635.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:07:37,849 INFO [optim.py:369] (0/4) 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 06:07:44,189 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62647.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:08:02,400 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62672.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:08:08,751 INFO [train.py:901] (0/4) Epoch 8, batch 6100, loss[loss=0.2175, simple_loss=0.2905, pruned_loss=0.07223, over 6014.00 frames. ], tot_loss[loss=0.2557, simple_loss=0.3255, pruned_loss=0.0929, over 1607240.99 frames. ], batch size: 13, lr: 9.43e-03, grad_scale: 8.0 +2023-02-06 06:08:10,156 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62684.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:08:21,655 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-02-06 06:08:37,233 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-06 06:08:43,079 INFO [train.py:901] (0/4) Epoch 8, batch 6150, loss[loss=0.2426, simple_loss=0.3255, pruned_loss=0.0798, over 8508.00 frames. ], tot_loss[loss=0.2566, simple_loss=0.3261, pruned_loss=0.09356, over 1611507.50 frames. ], batch size: 26, lr: 9.42e-03, grad_scale: 8.0 +2023-02-06 06:08:47,076 INFO [optim.py:369] (0/4) 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 06:08:55,074 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62750.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:08:57,034 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=62753.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:09:09,054 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.8186, 2.8133, 2.3042, 3.8244, 1.5479, 1.8373, 2.0479, 3.1056], + device='cuda:0'), covar=tensor([0.0666, 0.1070, 0.1095, 0.0275, 0.1468, 0.1766, 0.1423, 0.0834], + device='cuda:0'), in_proj_covar=tensor([0.0249, 0.0226, 0.0268, 0.0215, 0.0229, 0.0263, 0.0264, 0.0232], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 06:09:17,709 INFO [train.py:901] (0/4) Epoch 8, batch 6200, loss[loss=0.2445, simple_loss=0.3114, pruned_loss=0.08883, over 8469.00 frames. ], tot_loss[loss=0.2552, simple_loss=0.3249, pruned_loss=0.09278, over 1613511.49 frames. ], batch size: 25, lr: 9.42e-03, grad_scale: 16.0 +2023-02-06 06:09:23,750 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62791.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:09:52,744 INFO [train.py:901] (0/4) Epoch 8, batch 6250, loss[loss=0.2177, simple_loss=0.2988, pruned_loss=0.06826, over 8120.00 frames. ], tot_loss[loss=0.2534, simple_loss=0.3234, pruned_loss=0.09169, over 1615898.35 frames. ], batch size: 22, lr: 9.42e-03, grad_scale: 16.0 +2023-02-06 06:09:56,741 INFO [optim.py:369] (0/4) 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 06:09:57,784 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 +2023-02-06 06:10:08,418 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62855.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:10:16,900 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62868.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:10:24,983 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62880.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:10:26,036 INFO [train.py:901] (0/4) Epoch 8, batch 6300, loss[loss=0.2567, simple_loss=0.3377, pruned_loss=0.08779, over 7822.00 frames. ], tot_loss[loss=0.2521, simple_loss=0.3222, pruned_loss=0.091, over 1610378.88 frames. ], batch size: 20, lr: 9.41e-03, grad_scale: 16.0 +2023-02-06 06:10:28,871 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9653, 1.6973, 3.4689, 1.5815, 2.3097, 3.8841, 3.8188, 3.2738], + device='cuda:0'), covar=tensor([0.1049, 0.1308, 0.0339, 0.1811, 0.0916, 0.0228, 0.0478, 0.0626], + device='cuda:0'), in_proj_covar=tensor([0.0249, 0.0282, 0.0244, 0.0272, 0.0253, 0.0225, 0.0292, 0.0284], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 06:10:44,084 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=62907.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:10:44,437 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 +2023-02-06 06:10:50,363 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.96 vs. limit=5.0 +2023-02-06 06:11:01,296 INFO [train.py:901] (0/4) Epoch 8, batch 6350, loss[loss=0.2747, simple_loss=0.3314, pruned_loss=0.109, over 7927.00 frames. ], tot_loss[loss=0.2513, simple_loss=0.3216, pruned_loss=0.09046, over 1611539.76 frames. ], batch size: 20, lr: 9.41e-03, grad_scale: 16.0 +2023-02-06 06:11:05,345 INFO [optim.py:369] (0/4) 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 06:11:09,081 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.72 vs. limit=2.0 +2023-02-06 06:11:12,101 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=62947.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:11:12,141 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.6590, 5.7370, 5.0033, 2.0701, 5.0042, 5.3441, 5.2730, 4.7532], + device='cuda:0'), covar=tensor([0.0606, 0.0384, 0.0785, 0.4840, 0.0741, 0.0613, 0.0944, 0.0665], + device='cuda:0'), in_proj_covar=tensor([0.0420, 0.0334, 0.0350, 0.0443, 0.0349, 0.0329, 0.0341, 0.0291], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 06:11:20,472 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.68 vs. limit=5.0 +2023-02-06 06:11:35,235 INFO [train.py:901] (0/4) Epoch 8, batch 6400, loss[loss=0.3117, simple_loss=0.3723, pruned_loss=0.1255, over 8337.00 frames. ], tot_loss[loss=0.2515, simple_loss=0.3218, pruned_loss=0.09057, over 1616617.41 frames. ], batch size: 26, lr: 9.41e-03, grad_scale: 16.0 +2023-02-06 06:11:43,457 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3602, 2.2920, 3.1743, 2.0904, 2.8021, 3.4981, 3.3671, 3.1924], + device='cuda:0'), covar=tensor([0.0798, 0.1000, 0.0544, 0.1410, 0.0848, 0.0254, 0.0505, 0.0514], + device='cuda:0'), in_proj_covar=tensor([0.0250, 0.0279, 0.0241, 0.0271, 0.0249, 0.0222, 0.0290, 0.0283], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 06:11:52,108 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63006.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:12:03,482 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63022.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:12:06,107 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.5033, 1.3329, 4.6509, 1.8821, 4.0743, 3.8882, 4.1250, 4.1007], + device='cuda:0'), covar=tensor([0.0451, 0.4302, 0.0379, 0.2903, 0.0960, 0.0701, 0.0535, 0.0595], + device='cuda:0'), in_proj_covar=tensor([0.0430, 0.0542, 0.0524, 0.0496, 0.0555, 0.0466, 0.0472, 0.0527], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 06:12:07,444 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=63028.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:12:09,561 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63031.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:12:10,070 INFO [train.py:901] (0/4) Epoch 8, batch 6450, loss[loss=0.2281, simple_loss=0.3194, pruned_loss=0.06836, over 8285.00 frames. ], tot_loss[loss=0.251, simple_loss=0.3219, pruned_loss=0.0901, over 1614419.26 frames. ], batch size: 23, lr: 9.40e-03, grad_scale: 16.0 +2023-02-06 06:12:14,122 INFO [optim.py:369] (0/4) 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 06:12:16,481 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5422, 1.8950, 3.1849, 1.3028, 2.3370, 1.9580, 1.6853, 1.9892], + device='cuda:0'), covar=tensor([0.1570, 0.2139, 0.0555, 0.3378, 0.1259, 0.2373, 0.1542, 0.2087], + device='cuda:0'), in_proj_covar=tensor([0.0483, 0.0490, 0.0538, 0.0570, 0.0604, 0.0537, 0.0459, 0.0607], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:0') +2023-02-06 06:12:31,364 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63062.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:12:45,218 INFO [train.py:901] (0/4) Epoch 8, batch 6500, loss[loss=0.2892, simple_loss=0.3508, pruned_loss=0.1138, over 8026.00 frames. ], tot_loss[loss=0.2497, simple_loss=0.3205, pruned_loss=0.08944, over 1611557.02 frames. ], batch size: 22, lr: 9.40e-03, grad_scale: 16.0 +2023-02-06 06:13:03,585 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.03 vs. limit=5.0 +2023-02-06 06:13:14,195 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63124.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:13:20,105 INFO [train.py:901] (0/4) Epoch 8, batch 6550, loss[loss=0.2659, simple_loss=0.3467, pruned_loss=0.09256, over 8498.00 frames. ], tot_loss[loss=0.2512, simple_loss=0.3221, pruned_loss=0.09014, over 1616805.24 frames. ], batch size: 29, lr: 9.40e-03, grad_scale: 16.0 +2023-02-06 06:13:22,296 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=63135.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:13:24,223 INFO [optim.py:369] (0/4) 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 06:13:27,744 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63143.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:13:27,804 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5686, 2.0609, 4.4834, 1.1699, 3.0654, 2.3209, 1.4829, 2.8453], + device='cuda:0'), covar=tensor([0.1494, 0.2003, 0.0607, 0.3269, 0.1253, 0.2187, 0.1533, 0.1916], + device='cuda:0'), in_proj_covar=tensor([0.0479, 0.0487, 0.0533, 0.0566, 0.0600, 0.0535, 0.0458, 0.0603], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 06:13:29,194 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2604, 2.5085, 1.8670, 2.0649, 2.0085, 1.4371, 1.6730, 2.0025], + device='cuda:0'), covar=tensor([0.1257, 0.0325, 0.0852, 0.0479, 0.0633, 0.1262, 0.0906, 0.0741], + device='cuda:0'), in_proj_covar=tensor([0.0345, 0.0235, 0.0308, 0.0294, 0.0304, 0.0315, 0.0337, 0.0309], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 06:13:31,958 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63149.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:13:46,245 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7658, 1.6680, 1.9096, 1.3648, 1.0148, 1.9881, 0.2233, 1.3290], + device='cuda:0'), covar=tensor([0.2667, 0.1838, 0.0626, 0.2146, 0.5692, 0.0555, 0.3875, 0.1940], + device='cuda:0'), in_proj_covar=tensor([0.0153, 0.0155, 0.0091, 0.0204, 0.0243, 0.0095, 0.0159, 0.0158], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 06:13:49,407 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-06 06:13:54,672 INFO [train.py:901] (0/4) Epoch 8, batch 6600, loss[loss=0.2346, simple_loss=0.3134, pruned_loss=0.07795, over 8475.00 frames. ], tot_loss[loss=0.2527, simple_loss=0.3233, pruned_loss=0.09107, over 1613413.37 frames. ], batch size: 25, lr: 9.39e-03, grad_scale: 16.0 +2023-02-06 06:13:56,251 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.5787, 1.9907, 3.0898, 2.4440, 2.7065, 2.2969, 1.8024, 1.3348], + device='cuda:0'), covar=tensor([0.2884, 0.3413, 0.0811, 0.2108, 0.1644, 0.1687, 0.1461, 0.3660], + device='cuda:0'), in_proj_covar=tensor([0.0845, 0.0806, 0.0680, 0.0792, 0.0892, 0.0741, 0.0673, 0.0727], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 06:14:07,499 WARNING [train.py:1067] (0/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-06 06:14:25,323 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.45 vs. limit=5.0 +2023-02-06 06:14:29,745 INFO [train.py:901] (0/4) Epoch 8, batch 6650, loss[loss=0.2359, simple_loss=0.3198, pruned_loss=0.076, over 8470.00 frames. ], tot_loss[loss=0.2532, simple_loss=0.3236, pruned_loss=0.09144, over 1612982.48 frames. ], batch size: 27, lr: 9.39e-03, grad_scale: 16.0 +2023-02-06 06:14:33,637 INFO [optim.py:369] (0/4) 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 06:14:34,502 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63250.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:15:01,176 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63278.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:15:03,731 INFO [train.py:901] (0/4) Epoch 8, batch 6700, loss[loss=0.2333, simple_loss=0.3038, pruned_loss=0.08143, over 7810.00 frames. ], tot_loss[loss=0.2534, simple_loss=0.3237, pruned_loss=0.09156, over 1613450.88 frames. ], batch size: 20, lr: 9.38e-03, grad_scale: 16.0 +2023-02-06 06:15:19,254 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63303.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:15:29,352 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63318.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:15:34,401 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 +2023-02-06 06:15:38,588 INFO [train.py:901] (0/4) Epoch 8, batch 6750, loss[loss=0.2423, simple_loss=0.3184, pruned_loss=0.08308, over 8253.00 frames. ], tot_loss[loss=0.2529, simple_loss=0.3236, pruned_loss=0.09112, over 1611379.95 frames. ], batch size: 24, lr: 9.38e-03, grad_scale: 8.0 +2023-02-06 06:15:43,297 INFO [optim.py:369] (0/4) 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 06:15:46,209 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63343.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:16:13,012 INFO [train.py:901] (0/4) Epoch 8, batch 6800, loss[loss=0.2485, simple_loss=0.314, pruned_loss=0.09152, over 7974.00 frames. ], tot_loss[loss=0.2561, simple_loss=0.3266, pruned_loss=0.09285, over 1620705.85 frames. ], batch size: 21, lr: 9.38e-03, grad_scale: 8.0 +2023-02-06 06:16:19,053 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.6438, 5.6756, 4.9972, 2.4179, 5.0486, 5.3156, 5.2303, 4.8056], + device='cuda:0'), covar=tensor([0.0584, 0.0412, 0.0958, 0.4412, 0.0736, 0.0599, 0.0989, 0.0754], + device='cuda:0'), in_proj_covar=tensor([0.0436, 0.0343, 0.0360, 0.0451, 0.0358, 0.0335, 0.0349, 0.0299], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 06:16:20,977 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-06 06:16:24,459 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63399.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:16:42,711 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63424.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:16:47,937 INFO [train.py:901] (0/4) Epoch 8, batch 6850, loss[loss=0.2431, simple_loss=0.323, pruned_loss=0.08157, over 8242.00 frames. ], tot_loss[loss=0.2564, simple_loss=0.3268, pruned_loss=0.09298, over 1623605.48 frames. ], batch size: 22, lr: 9.37e-03, grad_scale: 8.0 +2023-02-06 06:16:52,292 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63438.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:16:52,781 INFO [optim.py:369] (0/4) 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 06:16:53,659 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63440.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:16:57,223 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9496, 3.8521, 2.4130, 2.5069, 2.9139, 1.7499, 2.6114, 2.8918], + device='cuda:0'), covar=tensor([0.1537, 0.0282, 0.0883, 0.0794, 0.0597, 0.1320, 0.0989, 0.0971], + device='cuda:0'), in_proj_covar=tensor([0.0349, 0.0237, 0.0309, 0.0295, 0.0306, 0.0315, 0.0336, 0.0312], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 06:17:11,112 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-06 06:17:22,529 INFO [train.py:901] (0/4) Epoch 8, batch 6900, loss[loss=0.2211, simple_loss=0.2968, pruned_loss=0.07269, over 7791.00 frames. ], tot_loss[loss=0.2555, simple_loss=0.3264, pruned_loss=0.09227, over 1625735.05 frames. ], batch size: 19, lr: 9.37e-03, grad_scale: 8.0 +2023-02-06 06:17:39,577 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63506.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:17:58,281 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63531.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:17:58,749 INFO [train.py:901] (0/4) Epoch 8, batch 6950, loss[loss=0.2459, simple_loss=0.3045, pruned_loss=0.09364, over 7983.00 frames. ], tot_loss[loss=0.2551, simple_loss=0.3256, pruned_loss=0.09229, over 1619660.79 frames. ], batch size: 21, lr: 9.37e-03, grad_scale: 8.0 +2023-02-06 06:18:03,563 INFO [optim.py:369] (0/4) 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 06:18:18,587 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-06 06:18:32,004 INFO [train.py:901] (0/4) Epoch 8, batch 7000, loss[loss=0.2875, simple_loss=0.3435, pruned_loss=0.1158, over 8134.00 frames. ], tot_loss[loss=0.2559, simple_loss=0.326, pruned_loss=0.09291, over 1619402.62 frames. ], batch size: 22, lr: 9.36e-03, grad_scale: 8.0 +2023-02-06 06:18:32,804 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=63583.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:18:36,312 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63587.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:18:48,864 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63604.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:19:00,965 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 +2023-02-06 06:19:08,074 INFO [train.py:901] (0/4) Epoch 8, batch 7050, loss[loss=0.2477, simple_loss=0.3289, pruned_loss=0.08326, over 8292.00 frames. ], tot_loss[loss=0.2542, simple_loss=0.3248, pruned_loss=0.09184, over 1611300.47 frames. ], batch size: 23, lr: 9.36e-03, grad_scale: 8.0 +2023-02-06 06:19:12,575 INFO [optim.py:369] (0/4) 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 06:19:38,210 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.73 vs. limit=5.0 +2023-02-06 06:19:42,433 INFO [train.py:901] (0/4) Epoch 8, batch 7100, loss[loss=0.2587, simple_loss=0.3274, pruned_loss=0.09494, over 8475.00 frames. ], tot_loss[loss=0.2539, simple_loss=0.3242, pruned_loss=0.0918, over 1608682.13 frames. ], batch size: 25, lr: 9.35e-03, grad_scale: 8.0 +2023-02-06 06:19:53,540 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63698.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:20:17,179 INFO [train.py:901] (0/4) Epoch 8, batch 7150, loss[loss=0.2446, simple_loss=0.3083, pruned_loss=0.0904, over 7802.00 frames. ], tot_loss[loss=0.2556, simple_loss=0.326, pruned_loss=0.09259, over 1614388.81 frames. ], batch size: 20, lr: 9.35e-03, grad_scale: 8.0 +2023-02-06 06:20:21,739 INFO [optim.py:369] (0/4) 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 06:20:25,962 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3157, 1.6127, 1.6965, 0.8600, 1.7176, 1.2663, 0.2518, 1.5394], + device='cuda:0'), covar=tensor([0.0226, 0.0157, 0.0124, 0.0230, 0.0166, 0.0471, 0.0405, 0.0130], + device='cuda:0'), in_proj_covar=tensor([0.0360, 0.0287, 0.0231, 0.0340, 0.0275, 0.0437, 0.0333, 0.0314], + device='cuda:0'), out_proj_covar=tensor([1.0887e-04, 8.4336e-05, 6.8109e-05, 1.0058e-04, 8.2936e-05, 1.4226e-04, + 1.0089e-04, 9.4080e-05], device='cuda:0') +2023-02-06 06:20:51,753 INFO [train.py:901] (0/4) Epoch 8, batch 7200, loss[loss=0.2717, simple_loss=0.3504, pruned_loss=0.09648, over 8347.00 frames. ], tot_loss[loss=0.2542, simple_loss=0.3248, pruned_loss=0.0918, over 1616734.44 frames. ], batch size: 26, lr: 9.35e-03, grad_scale: 8.0 +2023-02-06 06:20:51,831 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=63782.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:20:53,157 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=63784.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:21:07,239 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.3684, 3.0907, 3.5183, 2.0592, 1.9230, 3.4330, 0.6963, 2.1642], + device='cuda:0'), covar=tensor([0.2248, 0.1610, 0.0464, 0.3196, 0.5082, 0.0713, 0.5257, 0.2577], + device='cuda:0'), in_proj_covar=tensor([0.0155, 0.0153, 0.0090, 0.0199, 0.0242, 0.0095, 0.0158, 0.0156], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0003, 0.0002], + device='cuda:0') +2023-02-06 06:21:23,981 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63828.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:21:26,391 INFO [train.py:901] (0/4) Epoch 8, batch 7250, loss[loss=0.32, simple_loss=0.3694, pruned_loss=0.1353, over 8505.00 frames. ], tot_loss[loss=0.2541, simple_loss=0.3245, pruned_loss=0.0919, over 1612889.85 frames. ], batch size: 28, lr: 9.34e-03, grad_scale: 8.0 +2023-02-06 06:21:30,981 INFO [optim.py:369] (0/4) 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 06:21:41,873 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6483, 1.5917, 1.9417, 1.4509, 1.0613, 1.9615, 0.2667, 1.3259], + device='cuda:0'), covar=tensor([0.3014, 0.2103, 0.0482, 0.2153, 0.5651, 0.0667, 0.4268, 0.2082], + device='cuda:0'), in_proj_covar=tensor([0.0154, 0.0152, 0.0090, 0.0197, 0.0241, 0.0094, 0.0157, 0.0155], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0003, 0.0002], + device='cuda:0') +2023-02-06 06:21:47,995 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63863.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:21:49,574 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 +2023-02-06 06:22:00,268 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0109, 2.3200, 1.7489, 2.7229, 1.4669, 1.5441, 1.9866, 2.3851], + device='cuda:0'), covar=tensor([0.0750, 0.0770, 0.1047, 0.0388, 0.1034, 0.1421, 0.0902, 0.0721], + device='cuda:0'), in_proj_covar=tensor([0.0243, 0.0222, 0.0264, 0.0213, 0.0222, 0.0258, 0.0263, 0.0229], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0004, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 06:22:00,779 INFO [train.py:901] (0/4) Epoch 8, batch 7300, loss[loss=0.2875, simple_loss=0.3536, pruned_loss=0.1107, over 8352.00 frames. ], tot_loss[loss=0.2533, simple_loss=0.3236, pruned_loss=0.09148, over 1611322.64 frames. ], batch size: 24, lr: 9.34e-03, grad_scale: 8.0 +2023-02-06 06:22:12,928 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63897.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:22:14,320 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63899.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:22:33,994 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2710, 1.4793, 1.4956, 1.3076, 1.1000, 1.4560, 1.6585, 1.4409], + device='cuda:0'), covar=tensor([0.0477, 0.1233, 0.1863, 0.1431, 0.0595, 0.1531, 0.0687, 0.0637], + device='cuda:0'), in_proj_covar=tensor([0.0108, 0.0161, 0.0199, 0.0165, 0.0112, 0.0169, 0.0123, 0.0142], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0006, 0.0006], + device='cuda:0') +2023-02-06 06:22:36,699 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=63931.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:22:37,305 INFO [train.py:901] (0/4) Epoch 8, batch 7350, loss[loss=0.3294, simple_loss=0.3878, pruned_loss=0.1355, over 8605.00 frames. ], tot_loss[loss=0.2532, simple_loss=0.3236, pruned_loss=0.09143, over 1609197.44 frames. ], batch size: 49, lr: 9.34e-03, grad_scale: 8.0 +2023-02-06 06:22:38,872 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63934.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:22:42,246 INFO [optim.py:369] (0/4) 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 06:22:48,924 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=63948.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:22:53,925 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63954.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:23:05,931 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-06 06:23:10,850 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63979.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:23:12,545 INFO [train.py:901] (0/4) Epoch 8, batch 7400, loss[loss=0.2201, simple_loss=0.2937, pruned_loss=0.07324, over 8103.00 frames. ], tot_loss[loss=0.252, simple_loss=0.3221, pruned_loss=0.09095, over 1607975.88 frames. ], batch size: 21, lr: 9.33e-03, grad_scale: 8.0 +2023-02-06 06:23:24,821 WARNING [train.py:1067] (0/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-06 06:23:25,610 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-64000.pt +2023-02-06 06:23:29,609 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-06 06:23:48,438 INFO [train.py:901] (0/4) Epoch 8, batch 7450, loss[loss=0.2324, simple_loss=0.3222, pruned_loss=0.07127, over 8465.00 frames. ], tot_loss[loss=0.2533, simple_loss=0.3234, pruned_loss=0.09158, over 1612429.33 frames. ], batch size: 25, lr: 9.33e-03, grad_scale: 8.0 +2023-02-06 06:23:53,167 INFO [optim.py:369] (0/4) 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 06:23:54,693 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64041.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:23:58,108 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64046.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:24:05,487 WARNING [train.py:1067] (0/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-06 06:24:10,178 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64063.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:24:23,172 INFO [train.py:901] (0/4) Epoch 8, batch 7500, loss[loss=0.2582, simple_loss=0.3267, pruned_loss=0.09482, over 8480.00 frames. ], tot_loss[loss=0.2528, simple_loss=0.3233, pruned_loss=0.09109, over 1616778.88 frames. ], batch size: 29, lr: 9.33e-03, grad_scale: 8.0 +2023-02-06 06:24:34,552 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 +2023-02-06 06:24:53,421 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.90 vs. limit=5.0 +2023-02-06 06:24:57,461 INFO [train.py:901] (0/4) Epoch 8, batch 7550, loss[loss=0.3141, simple_loss=0.3729, pruned_loss=0.1276, over 8669.00 frames. ], tot_loss[loss=0.2563, simple_loss=0.3265, pruned_loss=0.09306, over 1618898.33 frames. ], batch size: 34, lr: 9.32e-03, grad_scale: 8.0 +2023-02-06 06:25:02,130 INFO [optim.py:369] (0/4) 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 06:25:07,026 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.2475, 1.8335, 1.9344, 1.7077, 1.2465, 1.8768, 2.1722, 1.9824], + device='cuda:0'), covar=tensor([0.0440, 0.1186, 0.1668, 0.1261, 0.0614, 0.1402, 0.0641, 0.0539], + device='cuda:0'), in_proj_covar=tensor([0.0108, 0.0162, 0.0200, 0.0164, 0.0113, 0.0169, 0.0123, 0.0143], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0006, 0.0006], + device='cuda:0') +2023-02-06 06:25:11,678 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64153.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:25:13,072 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64155.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:25:24,180 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=64172.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:25:28,956 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64178.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:25:30,997 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64180.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:25:32,163 INFO [train.py:901] (0/4) Epoch 8, batch 7600, loss[loss=0.2798, simple_loss=0.3538, pruned_loss=0.1029, over 8541.00 frames. ], tot_loss[loss=0.2543, simple_loss=0.3245, pruned_loss=0.09203, over 1614883.66 frames. ], batch size: 31, lr: 9.32e-03, grad_scale: 8.0 +2023-02-06 06:25:49,059 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 8, batch 7650, loss[loss=0.2682, simple_loss=0.3397, pruned_loss=0.09838, over 8451.00 frames. ], tot_loss[loss=0.2548, simple_loss=0.325, pruned_loss=0.09234, over 1617270.69 frames. ], batch size: 27, lr: 9.31e-03, grad_scale: 8.0 +2023-02-06 06:26:11,833 INFO [optim.py:369] (0/4) 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 06:26:19,350 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64250.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:26:38,168 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=64278.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:26:39,043 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4911, 1.9882, 3.1022, 1.2800, 2.2023, 1.8864, 1.7760, 1.8605], + device='cuda:0'), covar=tensor([0.1640, 0.1821, 0.0812, 0.3477, 0.1430, 0.2541, 0.1513, 0.2110], + device='cuda:0'), in_proj_covar=tensor([0.0475, 0.0486, 0.0527, 0.0562, 0.0601, 0.0538, 0.0457, 0.0601], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 06:26:40,901 INFO [train.py:901] (0/4) Epoch 8, batch 7700, loss[loss=0.2166, simple_loss=0.2868, pruned_loss=0.07322, over 7921.00 frames. ], tot_loss[loss=0.2563, simple_loss=0.3263, pruned_loss=0.09315, over 1619570.85 frames. ], batch size: 20, lr: 9.31e-03, grad_scale: 8.0 +2023-02-06 06:26:45,263 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64287.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:26:56,501 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64302.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:27:08,123 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64319.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:27:10,003 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64322.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:27:10,466 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-06 06:27:13,352 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64327.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:27:14,859 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.69 vs. limit=2.0 +2023-02-06 06:27:16,533 INFO [train.py:901] (0/4) Epoch 8, batch 7750, loss[loss=0.3104, simple_loss=0.3723, pruned_loss=0.1242, over 8110.00 frames. ], tot_loss[loss=0.2556, simple_loss=0.3259, pruned_loss=0.09262, over 1622575.58 frames. ], batch size: 23, lr: 9.31e-03, grad_scale: 8.0 +2023-02-06 06:27:17,264 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4309, 1.4899, 1.6037, 1.3474, 1.2545, 1.4767, 1.7575, 1.7368], + device='cuda:0'), covar=tensor([0.0496, 0.1269, 0.1833, 0.1418, 0.0639, 0.1614, 0.0708, 0.0614], + device='cuda:0'), in_proj_covar=tensor([0.0108, 0.0162, 0.0200, 0.0166, 0.0112, 0.0170, 0.0122, 0.0143], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0006, 0.0006], + device='cuda:0') +2023-02-06 06:27:21,028 INFO [optim.py:369] (0/4) 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 06:27:25,263 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64344.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:27:38,614 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64363.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:27:51,129 INFO [train.py:901] (0/4) Epoch 8, batch 7800, loss[loss=0.2021, simple_loss=0.2823, pruned_loss=0.06101, over 7538.00 frames. ], tot_loss[loss=0.2545, simple_loss=0.3253, pruned_loss=0.09187, over 1620011.04 frames. ], batch size: 18, lr: 9.30e-03, grad_scale: 8.0 +2023-02-06 06:27:53,236 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=64385.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:27:58,776 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 8, batch 7850, loss[loss=0.2792, simple_loss=0.3387, pruned_loss=0.1099, over 8362.00 frames. ], tot_loss[loss=0.2543, simple_loss=0.325, pruned_loss=0.09179, over 1620516.59 frames. ], batch size: 24, lr: 9.30e-03, grad_scale: 8.0 +2023-02-06 06:28:30,097 INFO [optim.py:369] (0/4) 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 06:28:58,102 INFO [train.py:901] (0/4) Epoch 8, batch 7900, loss[loss=0.2109, simple_loss=0.2824, pruned_loss=0.06975, over 7532.00 frames. ], tot_loss[loss=0.253, simple_loss=0.3241, pruned_loss=0.09096, over 1616067.75 frames. ], batch size: 18, lr: 9.30e-03, grad_scale: 8.0 +2023-02-06 06:28:58,897 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64483.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:29:10,215 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64500.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:29:13,087 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 +2023-02-06 06:29:32,366 INFO [train.py:901] (0/4) Epoch 8, batch 7950, loss[loss=0.2549, simple_loss=0.3216, pruned_loss=0.0941, over 8360.00 frames. ], tot_loss[loss=0.2515, simple_loss=0.3228, pruned_loss=0.09003, over 1618009.46 frames. ], batch size: 24, lr: 9.29e-03, grad_scale: 8.0 +2023-02-06 06:29:37,071 INFO [optim.py:369] (0/4) 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 06:29:40,088 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64543.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:29:56,713 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64568.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:30:03,487 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64578.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:30:06,037 INFO [train.py:901] (0/4) Epoch 8, batch 8000, loss[loss=0.2373, simple_loss=0.3163, pruned_loss=0.07914, over 8438.00 frames. ], tot_loss[loss=0.2517, simple_loss=0.3229, pruned_loss=0.09024, over 1613413.13 frames. ], batch size: 29, lr: 9.29e-03, grad_scale: 8.0 +2023-02-06 06:30:14,234 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=64594.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:30:20,521 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64607.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:30:39,931 INFO [train.py:901] (0/4) Epoch 8, batch 8050, loss[loss=0.2161, simple_loss=0.2918, pruned_loss=0.07014, over 7544.00 frames. ], tot_loss[loss=0.2509, simple_loss=0.322, pruned_loss=0.08989, over 1598075.51 frames. ], batch size: 18, lr: 9.29e-03, grad_scale: 8.0 +2023-02-06 06:30:44,635 INFO [optim.py:369] (0/4) 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 06:30:51,558 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64649.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:31:02,945 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/epoch-8.pt +2023-02-06 06:31:13,933 WARNING [train.py:1067] (0/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-06 06:31:17,607 INFO [train.py:901] (0/4) Epoch 9, batch 0, loss[loss=0.2973, simple_loss=0.3649, pruned_loss=0.1149, over 8449.00 frames. ], tot_loss[loss=0.2973, simple_loss=0.3649, pruned_loss=0.1149, over 8449.00 frames. ], batch size: 27, lr: 8.79e-03, grad_scale: 8.0 +2023-02-06 06:31:17,608 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 06:31:28,852 INFO [train.py:935] (0/4) Epoch 9, validation: loss=0.1983, simple_loss=0.2974, pruned_loss=0.04961, over 944034.00 frames. +2023-02-06 06:31:28,853 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6689MB +2023-02-06 06:31:29,663 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64666.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:31:35,186 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64674.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:31:43,423 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-06 06:31:56,869 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=64707.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:31:58,423 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64709.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:32:02,443 INFO [train.py:901] (0/4) Epoch 9, batch 50, loss[loss=0.2512, simple_loss=0.3377, pruned_loss=0.08239, over 8499.00 frames. ], tot_loss[loss=0.2583, simple_loss=0.3304, pruned_loss=0.09316, over 367886.81 frames. ], batch size: 26, lr: 8.79e-03, grad_scale: 8.0 +2023-02-06 06:32:06,618 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64721.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:32:16,338 WARNING [train.py:1067] (0/4) 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] (0/4) 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 06:32:31,577 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64756.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:32:31,743 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 +2023-02-06 06:32:36,071 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 9, batch 100, loss[loss=0.2232, simple_loss=0.289, pruned_loss=0.07872, over 7531.00 frames. ], tot_loss[loss=0.2568, simple_loss=0.3271, pruned_loss=0.09325, over 642141.14 frames. ], batch size: 18, lr: 8.78e-03, grad_scale: 8.0 +2023-02-06 06:32:41,522 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64770.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:32:42,101 WARNING [train.py:1067] (0/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-06 06:32:49,740 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64781.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:33:11,709 INFO [train.py:901] (0/4) Epoch 9, batch 150, loss[loss=0.2839, simple_loss=0.3387, pruned_loss=0.1145, over 7810.00 frames. ], tot_loss[loss=0.2562, simple_loss=0.3275, pruned_loss=0.09248, over 860820.17 frames. ], batch size: 20, lr: 8.78e-03, grad_scale: 8.0 +2023-02-06 06:33:16,731 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64822.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:33:20,054 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) Epoch 9, batch 200, loss[loss=0.3503, simple_loss=0.3909, pruned_loss=0.1548, over 8339.00 frames. ], tot_loss[loss=0.2559, simple_loss=0.3265, pruned_loss=0.09268, over 1029274.19 frames. ], batch size: 26, lr: 8.78e-03, grad_scale: 8.0 +2023-02-06 06:34:21,137 INFO [train.py:901] (0/4) Epoch 9, batch 250, loss[loss=0.2367, simple_loss=0.3154, pruned_loss=0.07902, over 8045.00 frames. ], tot_loss[loss=0.2542, simple_loss=0.3253, pruned_loss=0.09149, over 1159356.76 frames. ], batch size: 22, lr: 8.77e-03, grad_scale: 8.0 +2023-02-06 06:34:34,287 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-06 06:34:36,812 INFO [optim.py:369] (0/4) 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 06:34:39,015 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64942.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:34:42,838 WARNING [train.py:1067] (0/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-06 06:34:44,889 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=64951.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:34:54,123 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 +2023-02-06 06:34:54,262 INFO [train.py:901] (0/4) Epoch 9, batch 300, loss[loss=0.2183, simple_loss=0.2907, pruned_loss=0.07295, over 7937.00 frames. ], tot_loss[loss=0.2523, simple_loss=0.3238, pruned_loss=0.09036, over 1262001.09 frames. ], batch size: 20, lr: 8.77e-03, grad_scale: 8.0 +2023-02-06 06:34:54,462 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64965.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:35:11,911 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64990.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:35:15,047 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64994.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:35:26,441 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 9, batch 350, loss[loss=0.2467, simple_loss=0.3163, pruned_loss=0.08857, over 8086.00 frames. ], tot_loss[loss=0.2511, simple_loss=0.3225, pruned_loss=0.08988, over 1338769.37 frames. ], batch size: 21, lr: 8.77e-03, grad_scale: 8.0 +2023-02-06 06:35:46,491 INFO [optim.py:369] (0/4) 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 06:36:03,863 INFO [train.py:901] (0/4) Epoch 9, batch 400, loss[loss=0.1963, simple_loss=0.2671, pruned_loss=0.06276, over 7439.00 frames. ], tot_loss[loss=0.2528, simple_loss=0.3241, pruned_loss=0.09072, over 1409986.03 frames. ], batch size: 17, lr: 8.76e-03, grad_scale: 8.0 +2023-02-06 06:36:03,936 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=65065.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:36:04,755 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65066.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:36:09,395 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65073.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:36:12,869 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65078.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:36:30,519 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65103.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:36:33,060 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=65107.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:36:37,770 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=65114.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:36:38,396 INFO [train.py:901] (0/4) Epoch 9, batch 450, loss[loss=0.2742, simple_loss=0.3361, pruned_loss=0.1062, over 8374.00 frames. ], tot_loss[loss=0.2526, simple_loss=0.3235, pruned_loss=0.09087, over 1452696.21 frames. ], batch size: 49, lr: 8.76e-03, grad_scale: 8.0 +2023-02-06 06:36:46,201 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4422, 1.1598, 1.3720, 1.0980, 0.7863, 1.1457, 1.1598, 1.0697], + device='cuda:0'), covar=tensor([0.0538, 0.1386, 0.1819, 0.1450, 0.0608, 0.1633, 0.0687, 0.0655], + device='cuda:0'), in_proj_covar=tensor([0.0108, 0.0161, 0.0200, 0.0164, 0.0111, 0.0168, 0.0123, 0.0141], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0006, 0.0006], + device='cuda:0') +2023-02-06 06:36:46,210 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65125.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:36:56,307 INFO [optim.py:369] (0/4) 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 06:37:13,478 INFO [train.py:901] (0/4) Epoch 9, batch 500, loss[loss=0.2112, simple_loss=0.3013, pruned_loss=0.06054, over 8037.00 frames. ], tot_loss[loss=0.2518, simple_loss=0.3226, pruned_loss=0.09052, over 1490133.09 frames. ], batch size: 22, lr: 8.76e-03, grad_scale: 8.0 +2023-02-06 06:37:23,268 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65180.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:37:35,030 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65198.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:37:45,914 INFO [train.py:901] (0/4) Epoch 9, batch 550, loss[loss=0.2457, simple_loss=0.3195, pruned_loss=0.08599, over 7810.00 frames. ], tot_loss[loss=0.2526, simple_loss=0.3233, pruned_loss=0.09094, over 1516294.61 frames. ], batch size: 20, lr: 8.75e-03, grad_scale: 8.0 +2023-02-06 06:37:51,957 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65222.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:37:52,660 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65223.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:37:56,718 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65229.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:38:03,161 INFO [optim.py:369] (0/4) 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 06:38:21,259 INFO [train.py:901] (0/4) Epoch 9, batch 600, loss[loss=0.2227, simple_loss=0.2966, pruned_loss=0.07435, over 7653.00 frames. ], tot_loss[loss=0.251, simple_loss=0.3219, pruned_loss=0.09009, over 1535941.18 frames. ], batch size: 19, lr: 8.75e-03, grad_scale: 8.0 +2023-02-06 06:38:37,883 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7725, 1.7506, 2.1344, 1.5658, 1.1734, 2.2301, 0.2268, 1.2213], + device='cuda:0'), covar=tensor([0.2785, 0.1853, 0.0502, 0.3092, 0.4751, 0.0476, 0.4193, 0.2466], + device='cuda:0'), in_proj_covar=tensor([0.0155, 0.0153, 0.0091, 0.0201, 0.0242, 0.0095, 0.0156, 0.0156], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 06:38:38,361 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-06 06:38:46,504 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65303.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:38:54,303 INFO [train.py:901] (0/4) Epoch 9, batch 650, loss[loss=0.286, simple_loss=0.3531, pruned_loss=0.1094, over 7932.00 frames. ], tot_loss[loss=0.2512, simple_loss=0.3222, pruned_loss=0.09007, over 1550052.34 frames. ], batch size: 20, lr: 8.75e-03, grad_scale: 16.0 +2023-02-06 06:38:55,145 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2250, 1.7845, 4.2050, 1.9898, 2.3336, 4.7939, 4.5799, 4.1193], + device='cuda:0'), covar=tensor([0.1117, 0.1536, 0.0284, 0.1822, 0.1024, 0.0210, 0.0398, 0.0636], + device='cuda:0'), in_proj_covar=tensor([0.0251, 0.0282, 0.0241, 0.0273, 0.0254, 0.0224, 0.0296, 0.0282], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 06:38:59,102 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=65338.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:39:10,880 INFO [optim.py:369] (0/4) 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 06:39:17,067 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65347.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:39:29,085 INFO [train.py:901] (0/4) Epoch 9, batch 700, loss[loss=0.2358, simple_loss=0.322, pruned_loss=0.07484, over 8207.00 frames. ], tot_loss[loss=0.2487, simple_loss=0.3203, pruned_loss=0.08851, over 1565082.59 frames. ], batch size: 23, lr: 8.74e-03, grad_scale: 16.0 +2023-02-06 06:39:40,993 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65406.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:40:03,865 INFO [train.py:901] (0/4) Epoch 9, batch 750, loss[loss=0.2459, simple_loss=0.3232, pruned_loss=0.08425, over 8247.00 frames. ], tot_loss[loss=0.2494, simple_loss=0.3208, pruned_loss=0.089, over 1575805.28 frames. ], batch size: 24, lr: 8.74e-03, grad_scale: 16.0 +2023-02-06 06:40:05,337 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=65417.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:40:18,160 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65436.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:40:19,949 INFO [optim.py:369] (0/4) 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 06:40:21,291 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-06 06:40:30,021 WARNING [train.py:1067] (0/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-06 06:40:30,185 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65453.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:40:36,145 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65461.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:40:38,577 INFO [train.py:901] (0/4) Epoch 9, batch 800, loss[loss=0.2347, simple_loss=0.3154, pruned_loss=0.07698, over 8467.00 frames. ], tot_loss[loss=0.2502, simple_loss=0.3214, pruned_loss=0.0895, over 1585380.20 frames. ], batch size: 25, lr: 8.74e-03, grad_scale: 16.0 +2023-02-06 06:40:47,447 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65478.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:40:53,602 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65485.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:40:55,599 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2303, 1.7812, 1.8363, 1.6541, 1.2618, 1.7437, 1.9072, 1.8161], + device='cuda:0'), covar=tensor([0.0502, 0.0972, 0.1421, 0.1177, 0.0597, 0.1178, 0.0594, 0.0494], + device='cuda:0'), in_proj_covar=tensor([0.0107, 0.0159, 0.0199, 0.0164, 0.0111, 0.0168, 0.0122, 0.0142], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0006, 0.0006], + device='cuda:0') +2023-02-06 06:41:05,671 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65503.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:41:10,493 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65510.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:41:13,594 INFO [train.py:901] (0/4) Epoch 9, batch 850, loss[loss=0.2374, simple_loss=0.2969, pruned_loss=0.08892, over 7692.00 frames. ], tot_loss[loss=0.25, simple_loss=0.3209, pruned_loss=0.08952, over 1592998.24 frames. ], batch size: 18, lr: 8.73e-03, grad_scale: 16.0 +2023-02-06 06:41:25,164 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65532.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:41:29,558 INFO [optim.py:369] (0/4) 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 06:41:37,094 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65550.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 06:41:47,587 INFO [train.py:901] (0/4) Epoch 9, batch 900, loss[loss=0.2891, simple_loss=0.3594, pruned_loss=0.1094, over 8455.00 frames. ], tot_loss[loss=0.2497, simple_loss=0.3207, pruned_loss=0.08934, over 1597480.90 frames. ], batch size: 27, lr: 8.73e-03, grad_scale: 16.0 +2023-02-06 06:41:51,814 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.3423, 1.3599, 4.5070, 1.7552, 3.9891, 3.7561, 4.1160, 3.9078], + device='cuda:0'), covar=tensor([0.0467, 0.3838, 0.0389, 0.2989, 0.1094, 0.0769, 0.0447, 0.0604], + device='cuda:0'), in_proj_covar=tensor([0.0431, 0.0552, 0.0540, 0.0503, 0.0570, 0.0483, 0.0477, 0.0540], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 06:42:03,885 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9701, 1.5919, 1.7116, 1.2710, 1.0068, 1.4018, 1.6076, 1.7495], + device='cuda:0'), covar=tensor([0.0557, 0.1102, 0.1645, 0.1406, 0.0611, 0.1428, 0.0668, 0.0577], + device='cuda:0'), in_proj_covar=tensor([0.0107, 0.0159, 0.0198, 0.0163, 0.0111, 0.0167, 0.0122, 0.0142], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0006, 0.0006], + device='cuda:0') +2023-02-06 06:42:12,210 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3038, 1.9049, 2.8862, 2.2165, 2.4457, 2.1173, 1.6344, 1.2047], + device='cuda:0'), covar=tensor([0.2977, 0.3009, 0.0859, 0.1962, 0.1565, 0.1758, 0.1509, 0.3291], + device='cuda:0'), in_proj_covar=tensor([0.0850, 0.0814, 0.0699, 0.0804, 0.0891, 0.0754, 0.0687, 0.0727], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 06:42:12,365 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 +2023-02-06 06:42:23,181 INFO [train.py:901] (0/4) Epoch 9, batch 950, loss[loss=0.2649, simple_loss=0.336, pruned_loss=0.09693, over 8463.00 frames. ], tot_loss[loss=0.2496, simple_loss=0.3207, pruned_loss=0.08919, over 1601313.44 frames. ], batch size: 25, lr: 8.73e-03, grad_scale: 16.0 +2023-02-06 06:42:39,187 INFO [optim.py:369] (0/4) 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 06:42:44,682 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=65647.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:42:50,344 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 9, batch 1000, loss[loss=0.187, simple_loss=0.263, pruned_loss=0.05552, over 7934.00 frames. ], tot_loss[loss=0.25, simple_loss=0.3208, pruned_loss=0.08958, over 1604466.93 frames. ], batch size: 20, lr: 8.72e-03, grad_scale: 16.0 +2023-02-06 06:43:23,100 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-06 06:43:27,438 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65709.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:43:31,859 INFO [train.py:901] (0/4) Epoch 9, batch 1050, loss[loss=0.2433, simple_loss=0.3242, pruned_loss=0.08122, over 8491.00 frames. ], tot_loss[loss=0.2492, simple_loss=0.3205, pruned_loss=0.08892, over 1607306.28 frames. ], batch size: 29, lr: 8.72e-03, grad_scale: 16.0 +2023-02-06 06:43:35,708 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-06 06:43:44,958 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65734.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:43:46,903 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4308, 1.8644, 3.0879, 1.1498, 2.2022, 1.9467, 1.5768, 1.8290], + device='cuda:0'), covar=tensor([0.1930, 0.2113, 0.0866, 0.4179, 0.1646, 0.2936, 0.1946, 0.2409], + device='cuda:0'), in_proj_covar=tensor([0.0480, 0.0494, 0.0531, 0.0576, 0.0608, 0.0543, 0.0465, 0.0607], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 06:43:47,988 INFO [optim.py:369] (0/4) 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 06:44:03,478 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 9, batch 1100, loss[loss=0.2473, simple_loss=0.3083, pruned_loss=0.09315, over 7800.00 frames. ], tot_loss[loss=0.2476, simple_loss=0.3188, pruned_loss=0.08816, over 1602135.17 frames. ], batch size: 19, lr: 8.72e-03, grad_scale: 16.0 +2023-02-06 06:44:20,621 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65788.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:44:35,706 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-02-06 06:44:38,693 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65813.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:44:39,885 INFO [train.py:901] (0/4) Epoch 9, batch 1150, loss[loss=0.2649, simple_loss=0.3176, pruned_loss=0.1062, over 7426.00 frames. ], tot_loss[loss=0.2491, simple_loss=0.32, pruned_loss=0.08914, over 1603675.11 frames. ], batch size: 17, lr: 8.71e-03, grad_scale: 16.0 +2023-02-06 06:44:42,076 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.0301, 1.1061, 4.2320, 1.6473, 3.6840, 3.5288, 3.7554, 3.6881], + device='cuda:0'), covar=tensor([0.0582, 0.4294, 0.0483, 0.3136, 0.1216, 0.0933, 0.0625, 0.0703], + device='cuda:0'), in_proj_covar=tensor([0.0431, 0.0550, 0.0538, 0.0502, 0.0569, 0.0483, 0.0479, 0.0541], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 06:44:44,628 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-06 06:44:56,759 INFO [optim.py:369] (0/4) 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 06:45:06,289 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2736, 1.8374, 4.0923, 1.8400, 2.4025, 4.7076, 4.6782, 4.1282], + device='cuda:0'), covar=tensor([0.1097, 0.1475, 0.0308, 0.1909, 0.1085, 0.0215, 0.0343, 0.0545], + device='cuda:0'), in_proj_covar=tensor([0.0255, 0.0286, 0.0246, 0.0277, 0.0261, 0.0228, 0.0302, 0.0289], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 06:45:09,010 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65856.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 06:45:09,693 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.7374, 2.3981, 3.7081, 2.7950, 2.9544, 2.5766, 1.9176, 1.7519], + device='cuda:0'), covar=tensor([0.2954, 0.3370, 0.0888, 0.2137, 0.1851, 0.1597, 0.1468, 0.3762], + device='cuda:0'), in_proj_covar=tensor([0.0848, 0.0810, 0.0699, 0.0804, 0.0895, 0.0752, 0.0686, 0.0732], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 06:45:14,580 INFO [train.py:901] (0/4) Epoch 9, batch 1200, loss[loss=0.2994, simple_loss=0.3611, pruned_loss=0.1188, over 8699.00 frames. ], tot_loss[loss=0.2503, simple_loss=0.3211, pruned_loss=0.0898, over 1606298.10 frames. ], batch size: 39, lr: 8.71e-03, grad_scale: 16.0 +2023-02-06 06:45:33,707 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=65894.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 06:45:47,641 INFO [train.py:901] (0/4) Epoch 9, batch 1250, loss[loss=0.2406, simple_loss=0.3019, pruned_loss=0.08962, over 7193.00 frames. ], tot_loss[loss=0.2517, simple_loss=0.3223, pruned_loss=0.09052, over 1608459.87 frames. ], batch size: 16, lr: 8.71e-03, grad_scale: 16.0 +2023-02-06 06:45:57,331 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([0.7802, 0.7382, 0.8231, 0.7856, 0.5188, 0.8298, 0.0919, 0.7065], + device='cuda:0'), covar=tensor([0.2396, 0.1658, 0.0682, 0.1356, 0.3978, 0.0557, 0.3233, 0.1872], + device='cuda:0'), in_proj_covar=tensor([0.0156, 0.0154, 0.0092, 0.0204, 0.0245, 0.0095, 0.0159, 0.0157], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 06:46:05,023 INFO [optim.py:369] (0/4) 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 06:46:23,823 INFO [train.py:901] (0/4) Epoch 9, batch 1300, loss[loss=0.216, simple_loss=0.2948, pruned_loss=0.06861, over 8248.00 frames. ], tot_loss[loss=0.2513, simple_loss=0.3222, pruned_loss=0.09022, over 1609271.90 frames. ], batch size: 22, lr: 8.70e-03, grad_scale: 16.0 +2023-02-06 06:46:48,091 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-66000.pt +2023-02-06 06:46:55,300 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66009.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 06:46:59,214 INFO [train.py:901] (0/4) Epoch 9, batch 1350, loss[loss=0.2397, simple_loss=0.303, pruned_loss=0.08821, over 7931.00 frames. ], tot_loss[loss=0.2515, simple_loss=0.3226, pruned_loss=0.09018, over 1612283.19 frames. ], batch size: 20, lr: 8.70e-03, grad_scale: 8.0 +2023-02-06 06:47:01,402 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66018.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:47:06,410 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 +2023-02-06 06:47:17,555 INFO [optim.py:369] (0/4) 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 06:47:19,776 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66043.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:47:35,110 INFO [train.py:901] (0/4) Epoch 9, batch 1400, loss[loss=0.2378, simple_loss=0.3061, pruned_loss=0.08472, over 8071.00 frames. ], tot_loss[loss=0.2522, simple_loss=0.323, pruned_loss=0.0907, over 1613391.32 frames. ], batch size: 21, lr: 8.70e-03, grad_scale: 8.0 +2023-02-06 06:48:09,440 INFO [train.py:901] (0/4) Epoch 9, batch 1450, loss[loss=0.2237, simple_loss=0.2846, pruned_loss=0.08142, over 7233.00 frames. ], tot_loss[loss=0.2519, simple_loss=0.3227, pruned_loss=0.09054, over 1612305.65 frames. ], batch size: 16, lr: 8.69e-03, grad_scale: 8.0 +2023-02-06 06:48:12,162 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-06 06:48:21,005 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8734, 2.0064, 2.3940, 1.9138, 1.2562, 2.4698, 0.4006, 1.4868], + device='cuda:0'), covar=tensor([0.2780, 0.1891, 0.0458, 0.2106, 0.5807, 0.0491, 0.4660, 0.2566], + device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0151, 0.0090, 0.0197, 0.0240, 0.0092, 0.0154, 0.0152], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 06:48:26,155 INFO [optim.py:369] (0/4) 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 06:48:42,248 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66162.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:48:44,183 INFO [train.py:901] (0/4) Epoch 9, batch 1500, loss[loss=0.2612, simple_loss=0.3334, pruned_loss=0.09455, over 8315.00 frames. ], tot_loss[loss=0.2509, simple_loss=0.3222, pruned_loss=0.08974, over 1618754.19 frames. ], batch size: 25, lr: 8.69e-03, grad_scale: 8.0 +2023-02-06 06:48:52,885 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=66200.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 06:49:09,817 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 +2023-02-06 06:49:18,739 INFO [train.py:901] (0/4) Epoch 9, batch 1550, loss[loss=0.2871, simple_loss=0.359, pruned_loss=0.1076, over 8356.00 frames. ], tot_loss[loss=0.2512, simple_loss=0.3226, pruned_loss=0.08991, over 1622540.03 frames. ], batch size: 24, lr: 8.69e-03, grad_scale: 8.0 +2023-02-06 06:49:35,630 INFO [optim.py:369] (0/4) 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 06:49:53,211 INFO [train.py:901] (0/4) Epoch 9, batch 1600, loss[loss=0.2134, simple_loss=0.2923, pruned_loss=0.06723, over 7258.00 frames. ], tot_loss[loss=0.2518, simple_loss=0.3235, pruned_loss=0.09008, over 1626508.50 frames. ], batch size: 16, lr: 8.68e-03, grad_scale: 8.0 +2023-02-06 06:49:53,460 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66265.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 06:50:02,310 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66276.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:50:11,771 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66290.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 06:50:29,757 INFO [train.py:901] (0/4) Epoch 9, batch 1650, loss[loss=0.2734, simple_loss=0.3491, pruned_loss=0.0989, over 8420.00 frames. ], tot_loss[loss=0.2512, simple_loss=0.3228, pruned_loss=0.08983, over 1619614.38 frames. ], batch size: 49, lr: 8.68e-03, grad_scale: 8.0 +2023-02-06 06:50:29,942 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66315.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 06:50:35,446 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2205, 1.8555, 2.6018, 2.0632, 2.2660, 2.1394, 1.7324, 0.9634], + device='cuda:0'), covar=tensor([0.3312, 0.3443, 0.0961, 0.2065, 0.1667, 0.1803, 0.1731, 0.3392], + device='cuda:0'), in_proj_covar=tensor([0.0860, 0.0820, 0.0703, 0.0811, 0.0899, 0.0761, 0.0692, 0.0736], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 06:50:46,548 INFO [optim.py:369] (0/4) 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 06:50:49,638 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 +2023-02-06 06:50:54,222 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 +2023-02-06 06:51:03,465 INFO [train.py:901] (0/4) Epoch 9, batch 1700, loss[loss=0.2255, simple_loss=0.2979, pruned_loss=0.07655, over 7792.00 frames. ], tot_loss[loss=0.2504, simple_loss=0.3223, pruned_loss=0.08923, over 1614512.95 frames. ], batch size: 19, lr: 8.68e-03, grad_scale: 8.0 +2023-02-06 06:51:39,905 INFO [train.py:901] (0/4) Epoch 9, batch 1750, loss[loss=0.2304, simple_loss=0.3118, pruned_loss=0.07444, over 8349.00 frames. ], tot_loss[loss=0.2513, simple_loss=0.3231, pruned_loss=0.08977, over 1613028.00 frames. ], batch size: 24, lr: 8.67e-03, grad_scale: 8.0 +2023-02-06 06:51:57,456 INFO [optim.py:369] (0/4) 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 06:52:13,915 INFO [train.py:901] (0/4) Epoch 9, batch 1800, loss[loss=0.2739, simple_loss=0.3416, pruned_loss=0.1031, over 8466.00 frames. ], tot_loss[loss=0.2522, simple_loss=0.3234, pruned_loss=0.09054, over 1609663.99 frames. ], batch size: 27, lr: 8.67e-03, grad_scale: 8.0 +2023-02-06 06:52:42,174 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 9, batch 1850, loss[loss=0.2834, simple_loss=0.3498, pruned_loss=0.1085, over 8347.00 frames. ], tot_loss[loss=0.2501, simple_loss=0.3216, pruned_loss=0.08927, over 1614239.73 frames. ], batch size: 26, lr: 8.67e-03, grad_scale: 8.0 +2023-02-06 06:52:53,697 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=66522.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 06:53:05,510 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66539.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 06:53:05,995 INFO [optim.py:369] (0/4) 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 06:53:11,893 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.9140, 1.5166, 6.0125, 2.0677, 5.2501, 4.9845, 5.5428, 5.4193], + device='cuda:0'), covar=tensor([0.0357, 0.4166, 0.0246, 0.2999, 0.0835, 0.0618, 0.0390, 0.0447], + device='cuda:0'), in_proj_covar=tensor([0.0437, 0.0554, 0.0537, 0.0510, 0.0574, 0.0488, 0.0482, 0.0544], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 06:53:16,636 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 +2023-02-06 06:53:22,984 INFO [train.py:901] (0/4) Epoch 9, batch 1900, loss[loss=0.2591, simple_loss=0.3291, pruned_loss=0.09452, over 8581.00 frames. ], tot_loss[loss=0.2513, simple_loss=0.323, pruned_loss=0.08983, over 1616520.76 frames. ], batch size: 49, lr: 8.66e-03, grad_scale: 8.0 +2023-02-06 06:53:27,124 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66571.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 06:53:43,852 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66596.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 06:53:47,130 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-06 06:53:56,559 INFO [train.py:901] (0/4) Epoch 9, batch 1950, loss[loss=0.2325, simple_loss=0.3182, pruned_loss=0.07337, over 8488.00 frames. ], tot_loss[loss=0.2515, simple_loss=0.3229, pruned_loss=0.09012, over 1607669.90 frames. ], batch size: 26, lr: 8.66e-03, grad_scale: 8.0 +2023-02-06 06:53:58,597 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-06 06:54:00,023 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=66620.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:54:00,817 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66621.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:54:12,890 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66637.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 06:54:13,546 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3814, 1.7908, 3.3803, 1.1602, 2.2000, 1.8300, 1.4304, 2.1098], + device='cuda:0'), covar=tensor([0.1796, 0.2383, 0.0710, 0.3944, 0.1884, 0.2998, 0.1937, 0.2517], + device='cuda:0'), in_proj_covar=tensor([0.0480, 0.0492, 0.0535, 0.0569, 0.0610, 0.0540, 0.0466, 0.0604], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 06:54:14,619 INFO [optim.py:369] (0/4) 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 06:54:20,062 WARNING [train.py:1067] (0/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-06 06:54:32,101 INFO [train.py:901] (0/4) Epoch 9, batch 2000, loss[loss=0.1991, simple_loss=0.2835, pruned_loss=0.05737, over 8319.00 frames. ], tot_loss[loss=0.2495, simple_loss=0.3213, pruned_loss=0.08882, over 1610714.63 frames. ], batch size: 25, lr: 8.66e-03, grad_scale: 8.0 +2023-02-06 06:54:43,828 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-06 06:55:06,982 INFO [train.py:901] (0/4) Epoch 9, batch 2050, loss[loss=0.2405, simple_loss=0.3099, pruned_loss=0.08555, over 7237.00 frames. ], tot_loss[loss=0.2485, simple_loss=0.3203, pruned_loss=0.08836, over 1610800.42 frames. ], batch size: 16, lr: 8.65e-03, grad_scale: 8.0 +2023-02-06 06:55:20,379 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66735.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:55:23,653 INFO [optim.py:369] (0/4) 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 06:55:31,186 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66749.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 06:55:42,433 INFO [train.py:901] (0/4) Epoch 9, batch 2100, loss[loss=0.234, simple_loss=0.3007, pruned_loss=0.08361, over 7419.00 frames. ], tot_loss[loss=0.2489, simple_loss=0.3205, pruned_loss=0.08865, over 1613290.61 frames. ], batch size: 17, lr: 8.65e-03, grad_scale: 8.0 +2023-02-06 06:56:17,421 INFO [train.py:901] (0/4) Epoch 9, batch 2150, loss[loss=0.2327, simple_loss=0.3162, pruned_loss=0.07458, over 8101.00 frames. ], tot_loss[loss=0.25, simple_loss=0.3212, pruned_loss=0.08939, over 1610565.21 frames. ], batch size: 23, lr: 8.65e-03, grad_scale: 8.0 +2023-02-06 06:56:34,547 INFO [optim.py:369] (0/4) 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 06:56:36,817 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.49 vs. limit=5.0 +2023-02-06 06:56:53,074 INFO [train.py:901] (0/4) Epoch 9, batch 2200, loss[loss=0.288, simple_loss=0.3475, pruned_loss=0.1143, over 7720.00 frames. ], tot_loss[loss=0.2495, simple_loss=0.3216, pruned_loss=0.08866, over 1613441.52 frames. ], batch size: 77, lr: 8.64e-03, grad_scale: 8.0 +2023-02-06 06:57:01,059 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66877.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:57:05,631 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=66883.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 06:57:12,263 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66893.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 06:57:18,133 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66902.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:57:27,004 INFO [train.py:901] (0/4) Epoch 9, batch 2250, loss[loss=0.2373, simple_loss=0.3037, pruned_loss=0.08544, over 7928.00 frames. ], tot_loss[loss=0.2503, simple_loss=0.322, pruned_loss=0.08925, over 1615303.22 frames. ], batch size: 20, lr: 8.64e-03, grad_scale: 8.0 +2023-02-06 06:57:29,215 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66918.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 06:57:43,696 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 9, batch 2300, loss[loss=0.1896, simple_loss=0.2629, pruned_loss=0.05811, over 7541.00 frames. ], tot_loss[loss=0.2496, simple_loss=0.3214, pruned_loss=0.08894, over 1613843.26 frames. ], batch size: 18, lr: 8.64e-03, grad_scale: 8.0 +2023-02-06 06:58:04,398 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 +2023-02-06 06:58:07,524 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66975.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:58:19,746 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66991.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:58:24,438 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66998.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 06:58:36,115 INFO [train.py:901] (0/4) Epoch 9, batch 2350, loss[loss=0.2492, simple_loss=0.3294, pruned_loss=0.08446, over 8242.00 frames. ], tot_loss[loss=0.2497, simple_loss=0.3213, pruned_loss=0.08904, over 1613483.36 frames. ], batch size: 22, lr: 8.63e-03, grad_scale: 8.0 +2023-02-06 06:58:36,990 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67016.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:58:43,074 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5363, 1.7470, 1.7496, 1.2889, 1.8564, 1.3756, 0.8963, 1.5487], + device='cuda:0'), covar=tensor([0.0254, 0.0138, 0.0103, 0.0212, 0.0147, 0.0357, 0.0342, 0.0128], + device='cuda:0'), in_proj_covar=tensor([0.0366, 0.0292, 0.0243, 0.0354, 0.0283, 0.0443, 0.0336, 0.0320], + device='cuda:0'), out_proj_covar=tensor([1.0935e-04, 8.5246e-05, 7.1412e-05, 1.0370e-04, 8.4229e-05, 1.4228e-04, + 1.0067e-04, 9.4783e-05], device='cuda:0') +2023-02-06 06:58:53,558 INFO [optim.py:369] (0/4) 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 06:59:10,049 INFO [train.py:901] (0/4) Epoch 9, batch 2400, loss[loss=0.2569, simple_loss=0.3308, pruned_loss=0.09149, over 8332.00 frames. ], tot_loss[loss=0.2525, simple_loss=0.3232, pruned_loss=0.09091, over 1614204.94 frames. ], batch size: 26, lr: 8.63e-03, grad_scale: 8.0 +2023-02-06 06:59:10,298 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6519, 1.9679, 2.1436, 1.3709, 2.2829, 1.3721, 0.8238, 1.7887], + device='cuda:0'), covar=tensor([0.0397, 0.0198, 0.0158, 0.0297, 0.0162, 0.0557, 0.0471, 0.0178], + device='cuda:0'), in_proj_covar=tensor([0.0367, 0.0292, 0.0243, 0.0355, 0.0284, 0.0441, 0.0336, 0.0321], + device='cuda:0'), out_proj_covar=tensor([1.0949e-04, 8.5292e-05, 7.1180e-05, 1.0401e-04, 8.4612e-05, 1.4138e-04, + 1.0074e-04, 9.4993e-05], device='cuda:0') +2023-02-06 06:59:28,514 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=67093.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 06:59:34,522 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67101.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 06:59:44,442 INFO [train.py:901] (0/4) Epoch 9, batch 2450, loss[loss=0.265, simple_loss=0.3418, pruned_loss=0.09405, over 8555.00 frames. ], tot_loss[loss=0.2517, simple_loss=0.3226, pruned_loss=0.09035, over 1613102.19 frames. ], batch size: 31, lr: 8.63e-03, grad_scale: 8.0 +2023-02-06 07:00:02,012 INFO [optim.py:369] (0/4) 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 07:00:17,990 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8085, 2.0313, 1.6868, 2.5521, 1.3719, 1.2720, 1.7659, 2.0865], + device='cuda:0'), covar=tensor([0.0888, 0.1038, 0.1209, 0.0524, 0.1218, 0.1908, 0.1088, 0.0840], + device='cuda:0'), in_proj_covar=tensor([0.0247, 0.0226, 0.0265, 0.0218, 0.0224, 0.0264, 0.0267, 0.0231], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 07:00:18,498 INFO [train.py:901] (0/4) Epoch 9, batch 2500, loss[loss=0.2448, simple_loss=0.3222, pruned_loss=0.08372, over 8200.00 frames. ], tot_loss[loss=0.2523, simple_loss=0.3227, pruned_loss=0.09092, over 1609753.07 frames. ], batch size: 23, lr: 8.62e-03, grad_scale: 8.0 +2023-02-06 07:00:48,176 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67208.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 07:00:52,872 INFO [train.py:901] (0/4) Epoch 9, batch 2550, loss[loss=0.2323, simple_loss=0.3001, pruned_loss=0.08221, over 7545.00 frames. ], tot_loss[loss=0.2506, simple_loss=0.3218, pruned_loss=0.08973, over 1615470.74 frames. ], batch size: 18, lr: 8.62e-03, grad_scale: 8.0 +2023-02-06 07:01:00,606 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1486, 1.8167, 2.7658, 2.2177, 2.5190, 2.0609, 1.5612, 1.1070], + device='cuda:0'), covar=tensor([0.3543, 0.3461, 0.0972, 0.2151, 0.1587, 0.1886, 0.1641, 0.3814], + device='cuda:0'), in_proj_covar=tensor([0.0844, 0.0811, 0.0693, 0.0807, 0.0894, 0.0752, 0.0684, 0.0733], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 07:01:12,346 INFO [optim.py:369] (0/4) 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 07:01:21,880 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67254.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 07:01:26,653 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.43 vs. limit=5.0 +2023-02-06 07:01:29,611 INFO [train.py:901] (0/4) Epoch 9, batch 2600, loss[loss=0.2384, simple_loss=0.3131, pruned_loss=0.0819, over 8331.00 frames. ], tot_loss[loss=0.2522, simple_loss=0.323, pruned_loss=0.09067, over 1619530.42 frames. ], batch size: 26, lr: 8.62e-03, grad_scale: 8.0 +2023-02-06 07:01:39,251 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 9, batch 2650, loss[loss=0.2679, simple_loss=0.3448, pruned_loss=0.09557, over 8341.00 frames. ], tot_loss[loss=0.2518, simple_loss=0.3232, pruned_loss=0.09018, over 1620092.26 frames. ], batch size: 25, lr: 8.62e-03, grad_scale: 8.0 +2023-02-06 07:02:06,288 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=67319.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:02:21,821 INFO [optim.py:369] (0/4) 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 07:02:29,142 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 +2023-02-06 07:02:37,550 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4534, 2.0530, 2.1835, 1.3375, 2.2653, 1.4776, 0.6173, 1.7587], + device='cuda:0'), covar=tensor([0.0420, 0.0183, 0.0108, 0.0306, 0.0213, 0.0563, 0.0537, 0.0168], + device='cuda:0'), in_proj_covar=tensor([0.0370, 0.0293, 0.0246, 0.0359, 0.0285, 0.0446, 0.0340, 0.0324], + device='cuda:0'), out_proj_covar=tensor([1.1052e-04, 8.5331e-05, 7.2211e-05, 1.0526e-04, 8.4795e-05, 1.4292e-04, + 1.0181e-04, 9.6105e-05], device='cuda:0') +2023-02-06 07:02:38,164 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4323, 2.7561, 1.9344, 2.2077, 2.1012, 1.6083, 2.1019, 2.2203], + device='cuda:0'), covar=tensor([0.1157, 0.0291, 0.0880, 0.0548, 0.0583, 0.1162, 0.0860, 0.0818], + device='cuda:0'), in_proj_covar=tensor([0.0344, 0.0233, 0.0310, 0.0296, 0.0305, 0.0320, 0.0340, 0.0308], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 07:02:39,313 INFO [train.py:901] (0/4) Epoch 9, batch 2700, loss[loss=0.3204, simple_loss=0.3721, pruned_loss=0.1344, over 8474.00 frames. ], tot_loss[loss=0.2522, simple_loss=0.3237, pruned_loss=0.09042, over 1619084.51 frames. ], batch size: 49, lr: 8.61e-03, grad_scale: 8.0 +2023-02-06 07:03:09,421 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.1016, 2.2781, 1.6518, 1.8594, 1.7560, 1.4206, 1.5910, 1.7308], + device='cuda:0'), covar=tensor([0.1186, 0.0329, 0.0971, 0.0516, 0.0644, 0.1225, 0.0939, 0.0752], + device='cuda:0'), in_proj_covar=tensor([0.0348, 0.0236, 0.0313, 0.0298, 0.0307, 0.0321, 0.0342, 0.0309], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 07:03:13,103 INFO [train.py:901] (0/4) Epoch 9, batch 2750, loss[loss=0.2637, simple_loss=0.331, pruned_loss=0.09817, over 7915.00 frames. ], tot_loss[loss=0.2505, simple_loss=0.3222, pruned_loss=0.08944, over 1617455.36 frames. ], batch size: 20, lr: 8.61e-03, grad_scale: 8.0 +2023-02-06 07:03:26,069 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67434.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:03:29,963 INFO [optim.py:369] (0/4) 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 07:03:34,242 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=67445.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:03:47,998 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67464.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 07:03:48,379 INFO [train.py:901] (0/4) Epoch 9, batch 2800, loss[loss=0.2867, simple_loss=0.3537, pruned_loss=0.1099, over 8347.00 frames. ], tot_loss[loss=0.2504, simple_loss=0.3221, pruned_loss=0.0893, over 1617293.72 frames. ], batch size: 26, lr: 8.61e-03, grad_scale: 8.0 +2023-02-06 07:04:05,319 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67489.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 07:04:23,679 INFO [train.py:901] (0/4) Epoch 9, batch 2850, loss[loss=0.2172, simple_loss=0.2877, pruned_loss=0.07332, over 7796.00 frames. ], tot_loss[loss=0.251, simple_loss=0.3224, pruned_loss=0.08982, over 1612754.81 frames. ], batch size: 19, lr: 8.60e-03, grad_scale: 8.0 +2023-02-06 07:04:29,811 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.6842, 3.0910, 2.7213, 3.9680, 1.9375, 2.1736, 2.3765, 3.2578], + device='cuda:0'), covar=tensor([0.0758, 0.0756, 0.0951, 0.0318, 0.1256, 0.1533, 0.1242, 0.0850], + device='cuda:0'), in_proj_covar=tensor([0.0247, 0.0223, 0.0263, 0.0215, 0.0223, 0.0261, 0.0265, 0.0228], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 07:04:34,537 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67531.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:04:40,479 INFO [optim.py:369] (0/4) 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 07:04:55,075 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67560.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:04:58,190 INFO [train.py:901] (0/4) Epoch 9, batch 2900, loss[loss=0.234, simple_loss=0.3171, pruned_loss=0.07543, over 8460.00 frames. ], tot_loss[loss=0.2505, simple_loss=0.3221, pruned_loss=0.08941, over 1613418.60 frames. ], batch size: 27, lr: 8.60e-03, grad_scale: 8.0 +2023-02-06 07:05:04,081 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 +2023-02-06 07:05:24,281 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-06 07:05:33,897 INFO [train.py:901] (0/4) Epoch 9, batch 2950, loss[loss=0.2599, simple_loss=0.3424, pruned_loss=0.08872, over 8098.00 frames. ], tot_loss[loss=0.2496, simple_loss=0.322, pruned_loss=0.08861, over 1615759.99 frames. ], batch size: 23, lr: 8.60e-03, grad_scale: 8.0 +2023-02-06 07:05:51,259 INFO [optim.py:369] (0/4) 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 07:05:55,134 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.98 vs. limit=5.0 +2023-02-06 07:06:08,207 INFO [train.py:901] (0/4) Epoch 9, batch 3000, loss[loss=0.2396, simple_loss=0.3183, pruned_loss=0.08043, over 8026.00 frames. ], tot_loss[loss=0.2485, simple_loss=0.3207, pruned_loss=0.08817, over 1609904.87 frames. ], batch size: 22, lr: 8.59e-03, grad_scale: 8.0 +2023-02-06 07:06:08,208 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 07:06:20,344 INFO [train.py:935] (0/4) Epoch 9, validation: loss=0.1965, simple_loss=0.2957, pruned_loss=0.04864, over 944034.00 frames. +2023-02-06 07:06:20,346 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-06 07:06:37,426 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67690.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:06:43,344 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67698.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:06:52,168 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67710.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:06:55,499 INFO [train.py:901] (0/4) Epoch 9, batch 3050, loss[loss=0.2337, simple_loss=0.3167, pruned_loss=0.07532, over 8108.00 frames. ], tot_loss[loss=0.2488, simple_loss=0.3208, pruned_loss=0.08839, over 1612441.01 frames. ], batch size: 23, lr: 8.59e-03, grad_scale: 8.0 +2023-02-06 07:06:55,703 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67715.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:07:00,566 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3283, 1.1279, 1.4564, 1.1604, 0.7564, 1.2304, 1.1856, 1.0669], + device='cuda:0'), covar=tensor([0.0548, 0.1343, 0.1723, 0.1394, 0.0549, 0.1548, 0.0669, 0.0664], + device='cuda:0'), in_proj_covar=tensor([0.0105, 0.0158, 0.0196, 0.0161, 0.0108, 0.0166, 0.0121, 0.0140], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0005, 0.0008, 0.0006, 0.0006], + device='cuda:0') +2023-02-06 07:07:13,228 INFO [optim.py:369] (0/4) 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 07:07:29,789 INFO [train.py:901] (0/4) Epoch 9, batch 3100, loss[loss=0.2455, simple_loss=0.3153, pruned_loss=0.08778, over 7820.00 frames. ], tot_loss[loss=0.2493, simple_loss=0.321, pruned_loss=0.08885, over 1614085.66 frames. ], batch size: 20, lr: 8.59e-03, grad_scale: 8.0 +2023-02-06 07:07:57,805 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3738, 1.4976, 1.7187, 1.3891, 1.0799, 1.4836, 1.8110, 1.7428], + device='cuda:0'), covar=tensor([0.0450, 0.1278, 0.1614, 0.1337, 0.0559, 0.1449, 0.0699, 0.0573], + device='cuda:0'), in_proj_covar=tensor([0.0106, 0.0159, 0.0197, 0.0162, 0.0109, 0.0167, 0.0121, 0.0141], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0006, 0.0006], + device='cuda:0') +2023-02-06 07:08:04,167 INFO [train.py:901] (0/4) Epoch 9, batch 3150, loss[loss=0.2543, simple_loss=0.3205, pruned_loss=0.09407, over 8504.00 frames. ], tot_loss[loss=0.2498, simple_loss=0.3212, pruned_loss=0.08919, over 1612524.39 frames. ], batch size: 28, lr: 8.58e-03, grad_scale: 8.0 +2023-02-06 07:08:05,040 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67816.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:08:21,137 INFO [optim.py:369] (0/4) 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 07:08:21,976 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67841.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:08:37,976 INFO [train.py:901] (0/4) Epoch 9, batch 3200, loss[loss=0.1988, simple_loss=0.271, pruned_loss=0.06331, over 7929.00 frames. ], tot_loss[loss=0.2516, simple_loss=0.3227, pruned_loss=0.09026, over 1611919.19 frames. ], batch size: 20, lr: 8.58e-03, grad_scale: 8.0 +2023-02-06 07:08:38,989 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-02-06 07:08:44,726 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=67875.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:08:47,156 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 +2023-02-06 07:09:12,191 INFO [train.py:901] (0/4) Epoch 9, batch 3250, loss[loss=0.2436, simple_loss=0.3134, pruned_loss=0.0869, over 7704.00 frames. ], tot_loss[loss=0.2494, simple_loss=0.3212, pruned_loss=0.08875, over 1611093.99 frames. ], batch size: 18, lr: 8.58e-03, grad_scale: 8.0 +2023-02-06 07:09:14,078 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 +2023-02-06 07:09:29,470 INFO [optim.py:369] (0/4) 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 07:09:31,093 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6992, 2.2112, 3.4610, 1.2219, 2.5731, 2.1019, 1.7887, 2.3423], + device='cuda:0'), covar=tensor([0.1664, 0.1920, 0.0800, 0.3774, 0.1477, 0.2624, 0.1667, 0.2235], + device='cuda:0'), in_proj_covar=tensor([0.0479, 0.0494, 0.0527, 0.0566, 0.0601, 0.0540, 0.0460, 0.0603], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 07:09:46,666 INFO [train.py:901] (0/4) Epoch 9, batch 3300, loss[loss=0.2651, simple_loss=0.3409, pruned_loss=0.09465, over 8349.00 frames. ], tot_loss[loss=0.2509, simple_loss=0.3226, pruned_loss=0.08962, over 1611753.28 frames. ], batch size: 26, lr: 8.57e-03, grad_scale: 8.0 +2023-02-06 07:10:04,398 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67999.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:10:11,136 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-68000.pt +2023-02-06 07:10:22,541 INFO [train.py:901] (0/4) Epoch 9, batch 3350, loss[loss=0.2395, simple_loss=0.2968, pruned_loss=0.09113, over 7791.00 frames. ], tot_loss[loss=0.2513, simple_loss=0.3232, pruned_loss=0.08966, over 1614344.05 frames. ], batch size: 19, lr: 8.57e-03, grad_scale: 16.0 +2023-02-06 07:10:39,228 INFO [optim.py:369] (0/4) 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 07:10:40,596 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=68042.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:10:49,086 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=68054.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:10:56,400 INFO [train.py:901] (0/4) Epoch 9, batch 3400, loss[loss=0.2511, simple_loss=0.3012, pruned_loss=0.1005, over 5973.00 frames. ], tot_loss[loss=0.2501, simple_loss=0.3222, pruned_loss=0.08901, over 1614535.08 frames. ], batch size: 13, lr: 8.57e-03, grad_scale: 16.0 +2023-02-06 07:11:24,318 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68105.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 07:11:30,817 INFO [train.py:901] (0/4) Epoch 9, batch 3450, loss[loss=0.195, simple_loss=0.2629, pruned_loss=0.06353, over 7666.00 frames. ], tot_loss[loss=0.2515, simple_loss=0.3231, pruned_loss=0.08997, over 1616209.89 frames. ], batch size: 19, lr: 8.56e-03, grad_scale: 16.0 +2023-02-06 07:11:46,275 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0766, 2.4282, 1.8999, 2.8631, 1.5527, 1.5102, 1.9456, 2.4329], + device='cuda:0'), covar=tensor([0.0851, 0.0838, 0.1073, 0.0423, 0.1181, 0.1676, 0.1160, 0.0777], + device='cuda:0'), in_proj_covar=tensor([0.0244, 0.0225, 0.0263, 0.0219, 0.0223, 0.0262, 0.0266, 0.0227], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 07:11:48,139 INFO [optim.py:369] (0/4) 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 07:11:59,832 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68157.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:12:05,797 INFO [train.py:901] (0/4) Epoch 9, batch 3500, loss[loss=0.3393, simple_loss=0.3907, pruned_loss=0.144, over 6962.00 frames. ], tot_loss[loss=0.2527, simple_loss=0.3239, pruned_loss=0.09074, over 1613872.14 frames. ], batch size: 71, lr: 8.56e-03, grad_scale: 8.0 +2023-02-06 07:12:08,690 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68169.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:12:17,955 WARNING [train.py:1067] (0/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-06 07:12:40,955 INFO [train.py:901] (0/4) Epoch 9, batch 3550, loss[loss=0.2362, simple_loss=0.3113, pruned_loss=0.08052, over 8140.00 frames. ], tot_loss[loss=0.2508, simple_loss=0.3224, pruned_loss=0.08958, over 1614619.63 frames. ], batch size: 22, lr: 8.56e-03, grad_scale: 8.0 +2023-02-06 07:12:58,953 INFO [optim.py:369] (0/4) 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] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68246.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:13:14,899 INFO [train.py:901] (0/4) Epoch 9, batch 3600, loss[loss=0.278, simple_loss=0.3522, pruned_loss=0.1019, over 8250.00 frames. ], tot_loss[loss=0.248, simple_loss=0.3199, pruned_loss=0.08802, over 1614199.37 frames. ], batch size: 24, lr: 8.56e-03, grad_scale: 8.0 +2023-02-06 07:13:19,101 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 9, batch 3650, loss[loss=0.2586, simple_loss=0.3317, pruned_loss=0.09272, over 8180.00 frames. ], tot_loss[loss=0.2471, simple_loss=0.3193, pruned_loss=0.08746, over 1613575.54 frames. ], batch size: 23, lr: 8.55e-03, grad_scale: 8.0 +2023-02-06 07:14:08,231 INFO [optim.py:369] (0/4) 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 07:14:09,680 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=68343.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:14:18,952 WARNING [train.py:1067] (0/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-06 07:14:25,007 INFO [train.py:901] (0/4) Epoch 9, batch 3700, loss[loss=0.2945, simple_loss=0.3676, pruned_loss=0.1107, over 8670.00 frames. ], tot_loss[loss=0.2475, simple_loss=0.3194, pruned_loss=0.08777, over 1612646.43 frames. ], batch size: 34, lr: 8.55e-03, grad_scale: 8.0 +2023-02-06 07:14:44,966 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.9842, 1.8321, 1.8275, 1.7063, 1.4407, 1.8456, 2.4051, 2.5769], + device='cuda:0'), covar=tensor([0.0461, 0.1115, 0.1566, 0.1240, 0.0493, 0.1338, 0.0554, 0.0453], + device='cuda:0'), in_proj_covar=tensor([0.0106, 0.0157, 0.0195, 0.0159, 0.0108, 0.0165, 0.0119, 0.0139], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0008, 0.0008, 0.0005, 0.0008, 0.0006, 0.0006], + device='cuda:0') +2023-02-06 07:14:57,586 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68413.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:14:58,733 INFO [train.py:901] (0/4) Epoch 9, batch 3750, loss[loss=0.2545, simple_loss=0.3361, pruned_loss=0.08641, over 8184.00 frames. ], tot_loss[loss=0.2483, simple_loss=0.3202, pruned_loss=0.08817, over 1617183.99 frames. ], batch size: 23, lr: 8.55e-03, grad_scale: 8.0 +2023-02-06 07:15:00,172 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68417.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:15:06,058 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68425.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:15:14,901 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68438.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:15:16,803 INFO [optim.py:369] (0/4) 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 07:15:22,893 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=68449.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 07:15:23,624 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68450.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:15:29,204 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68458.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:15:33,815 INFO [train.py:901] (0/4) Epoch 9, batch 3800, loss[loss=0.1881, simple_loss=0.2779, pruned_loss=0.04913, over 7814.00 frames. ], tot_loss[loss=0.2465, simple_loss=0.3183, pruned_loss=0.08731, over 1611893.02 frames. ], batch size: 20, lr: 8.54e-03, grad_scale: 8.0 +2023-02-06 07:15:37,786 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 +2023-02-06 07:16:07,819 INFO [train.py:901] (0/4) Epoch 9, batch 3850, loss[loss=0.2475, simple_loss=0.324, pruned_loss=0.08551, over 8258.00 frames. ], tot_loss[loss=0.2476, simple_loss=0.3196, pruned_loss=0.08785, over 1615960.52 frames. ], batch size: 24, lr: 8.54e-03, grad_scale: 8.0 +2023-02-06 07:16:25,023 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-06 07:16:25,655 INFO [optim.py:369] (0/4) 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 07:16:42,279 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68564.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 07:16:42,694 INFO [train.py:901] (0/4) Epoch 9, batch 3900, loss[loss=0.3332, simple_loss=0.399, pruned_loss=0.1338, over 8508.00 frames. ], tot_loss[loss=0.2487, simple_loss=0.3205, pruned_loss=0.0884, over 1616423.60 frames. ], batch size: 26, lr: 8.54e-03, grad_scale: 8.0 +2023-02-06 07:17:04,307 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68596.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:17:17,701 INFO [train.py:901] (0/4) Epoch 9, batch 3950, loss[loss=0.2115, simple_loss=0.2913, pruned_loss=0.06587, over 7655.00 frames. ], tot_loss[loss=0.248, simple_loss=0.3197, pruned_loss=0.08816, over 1613721.28 frames. ], batch size: 19, lr: 8.53e-03, grad_scale: 8.0 +2023-02-06 07:17:35,330 INFO [optim.py:369] (0/4) 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 07:17:51,772 INFO [train.py:901] (0/4) Epoch 9, batch 4000, loss[loss=0.2349, simple_loss=0.3214, pruned_loss=0.0742, over 8021.00 frames. ], tot_loss[loss=0.2462, simple_loss=0.3183, pruned_loss=0.08703, over 1611922.39 frames. ], batch size: 22, lr: 8.53e-03, grad_scale: 8.0 +2023-02-06 07:18:17,927 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68703.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:18:25,235 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68714.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:18:25,678 INFO [train.py:901] (0/4) Epoch 9, batch 4050, loss[loss=0.2046, simple_loss=0.2788, pruned_loss=0.06521, over 7549.00 frames. ], tot_loss[loss=0.2481, simple_loss=0.3199, pruned_loss=0.08812, over 1614859.88 frames. ], batch size: 18, lr: 8.53e-03, grad_scale: 8.0 +2023-02-06 07:18:41,426 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.64 vs. limit=2.0 +2023-02-06 07:18:42,621 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68739.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:18:43,699 INFO [optim.py:369] (0/4) 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 07:18:57,741 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=68761.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:18:58,668 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.37 vs. limit=5.0 +2023-02-06 07:19:00,371 INFO [train.py:901] (0/4) Epoch 9, batch 4100, loss[loss=0.2727, simple_loss=0.3429, pruned_loss=0.1013, over 8255.00 frames. ], tot_loss[loss=0.2473, simple_loss=0.3194, pruned_loss=0.08763, over 1608424.64 frames. ], batch size: 24, lr: 8.52e-03, grad_scale: 8.0 +2023-02-06 07:19:19,602 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2139, 2.2501, 1.5504, 1.9745, 1.7973, 1.2345, 1.6907, 1.7489], + device='cuda:0'), covar=tensor([0.1132, 0.0305, 0.1017, 0.0484, 0.0579, 0.1363, 0.0785, 0.0761], + device='cuda:0'), in_proj_covar=tensor([0.0351, 0.0238, 0.0314, 0.0301, 0.0307, 0.0324, 0.0348, 0.0317], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 07:19:23,693 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 +2023-02-06 07:19:34,807 INFO [train.py:901] (0/4) Epoch 9, batch 4150, loss[loss=0.2223, simple_loss=0.2916, pruned_loss=0.07649, over 7520.00 frames. ], tot_loss[loss=0.2483, simple_loss=0.3202, pruned_loss=0.08823, over 1607469.64 frames. ], batch size: 18, lr: 8.52e-03, grad_scale: 8.0 +2023-02-06 07:19:38,342 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68820.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 07:19:52,098 INFO [optim.py:369] (0/4) 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 07:19:55,811 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68845.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 07:20:02,869 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-06 07:20:09,486 INFO [train.py:901] (0/4) Epoch 9, batch 4200, loss[loss=0.2229, simple_loss=0.3099, pruned_loss=0.06801, over 8460.00 frames. ], tot_loss[loss=0.2477, simple_loss=0.3201, pruned_loss=0.08765, over 1609387.34 frames. ], batch size: 25, lr: 8.52e-03, grad_scale: 8.0 +2023-02-06 07:20:12,587 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 +2023-02-06 07:20:17,156 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68876.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:20:23,068 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-06 07:20:44,846 INFO [train.py:901] (0/4) Epoch 9, batch 4250, loss[loss=0.2416, simple_loss=0.3161, pruned_loss=0.08353, over 8361.00 frames. ], tot_loss[loss=0.2472, simple_loss=0.32, pruned_loss=0.08723, over 1610598.94 frames. ], batch size: 24, lr: 8.52e-03, grad_scale: 8.0 +2023-02-06 07:20:45,581 WARNING [train.py:1067] (0/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-06 07:20:51,294 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4099, 1.9114, 3.0770, 2.3874, 2.7730, 2.0754, 1.7306, 1.3874], + device='cuda:0'), covar=tensor([0.3192, 0.3464, 0.0905, 0.2035, 0.1467, 0.1852, 0.1521, 0.3646], + device='cuda:0'), in_proj_covar=tensor([0.0854, 0.0816, 0.0698, 0.0808, 0.0902, 0.0761, 0.0686, 0.0733], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 07:21:02,849 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=68940.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:21:03,456 INFO [optim.py:369] (0/4) 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 07:21:20,748 INFO [train.py:901] (0/4) Epoch 9, batch 4300, loss[loss=0.203, simple_loss=0.2933, pruned_loss=0.05636, over 8242.00 frames. ], tot_loss[loss=0.2484, simple_loss=0.3206, pruned_loss=0.08807, over 1611009.25 frames. ], batch size: 22, lr: 8.51e-03, grad_scale: 8.0 +2023-02-06 07:21:55,042 INFO [train.py:901] (0/4) Epoch 9, batch 4350, loss[loss=0.1942, simple_loss=0.2757, pruned_loss=0.0564, over 7980.00 frames. ], tot_loss[loss=0.2477, simple_loss=0.3202, pruned_loss=0.08759, over 1612913.52 frames. ], batch size: 21, lr: 8.51e-03, grad_scale: 8.0 +2023-02-06 07:22:03,359 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 +2023-02-06 07:22:12,980 INFO [optim.py:369] (0/4) 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 07:22:16,276 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-06 07:22:17,163 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1826, 1.5793, 1.5891, 1.4019, 1.2413, 1.3973, 1.7938, 1.7694], + device='cuda:0'), covar=tensor([0.0492, 0.1178, 0.1724, 0.1334, 0.0556, 0.1463, 0.0669, 0.0576], + device='cuda:0'), in_proj_covar=tensor([0.0106, 0.0159, 0.0197, 0.0161, 0.0110, 0.0167, 0.0122, 0.0141], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0006, 0.0006], + device='cuda:0') +2023-02-06 07:22:17,702 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=69047.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:22:23,135 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69055.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:22:27,146 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 9, batch 4400, loss[loss=0.2056, simple_loss=0.2695, pruned_loss=0.07089, over 7457.00 frames. ], tot_loss[loss=0.2476, simple_loss=0.3206, pruned_loss=0.08731, over 1617766.27 frames. ], batch size: 17, lr: 8.51e-03, grad_scale: 8.0 +2023-02-06 07:22:55,804 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-06 07:23:04,525 INFO [train.py:901] (0/4) Epoch 9, batch 4450, loss[loss=0.2025, simple_loss=0.2802, pruned_loss=0.06236, over 7430.00 frames. ], tot_loss[loss=0.2456, simple_loss=0.3185, pruned_loss=0.08631, over 1613368.94 frames. ], batch size: 17, lr: 8.50e-03, grad_scale: 8.0 +2023-02-06 07:23:16,664 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 07:23:32,401 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.1376, 4.1257, 3.7277, 1.7561, 3.6582, 3.7074, 3.7839, 3.3926], + device='cuda:0'), covar=tensor([0.0867, 0.0569, 0.1002, 0.4778, 0.0827, 0.1001, 0.1161, 0.0971], + device='cuda:0'), in_proj_covar=tensor([0.0442, 0.0350, 0.0369, 0.0459, 0.0363, 0.0345, 0.0362, 0.0311], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 07:23:33,153 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69157.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:23:37,252 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69162.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:23:39,006 INFO [train.py:901] (0/4) Epoch 9, batch 4500, loss[loss=0.2164, simple_loss=0.3058, pruned_loss=0.06353, over 8192.00 frames. ], tot_loss[loss=0.2466, simple_loss=0.3194, pruned_loss=0.08694, over 1614397.43 frames. ], batch size: 23, lr: 8.50e-03, grad_scale: 8.0 +2023-02-06 07:23:49,160 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-06 07:23:49,294 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69180.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:23:53,172 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69186.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:23:54,609 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8576, 2.1172, 1.6435, 2.6683, 1.0676, 1.5030, 1.7996, 2.1680], + device='cuda:0'), covar=tensor([0.0868, 0.0784, 0.1204, 0.0384, 0.1280, 0.1357, 0.0903, 0.0748], + device='cuda:0'), in_proj_covar=tensor([0.0250, 0.0225, 0.0266, 0.0223, 0.0228, 0.0264, 0.0268, 0.0230], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 07:24:13,186 INFO [train.py:901] (0/4) Epoch 9, batch 4550, loss[loss=0.2211, simple_loss=0.3118, pruned_loss=0.06517, over 8253.00 frames. ], tot_loss[loss=0.2464, simple_loss=0.3191, pruned_loss=0.08682, over 1615885.13 frames. ], batch size: 24, lr: 8.50e-03, grad_scale: 8.0 +2023-02-06 07:24:19,720 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4942, 1.6703, 2.8381, 1.1915, 2.0407, 1.7774, 1.5011, 1.7827], + device='cuda:0'), covar=tensor([0.1645, 0.2047, 0.0646, 0.3735, 0.1406, 0.2689, 0.1724, 0.2055], + device='cuda:0'), in_proj_covar=tensor([0.0475, 0.0491, 0.0522, 0.0564, 0.0601, 0.0542, 0.0462, 0.0603], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 07:24:31,949 INFO [optim.py:369] (0/4) 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 07:24:33,334 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69243.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:24:47,696 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69263.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 07:24:48,919 INFO [train.py:901] (0/4) Epoch 9, batch 4600, loss[loss=0.236, simple_loss=0.3153, pruned_loss=0.07834, over 8287.00 frames. ], tot_loss[loss=0.2464, simple_loss=0.3188, pruned_loss=0.08697, over 1613187.75 frames. ], batch size: 23, lr: 8.49e-03, grad_scale: 8.0 +2023-02-06 07:24:57,779 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.7622, 3.7629, 3.3713, 1.6352, 3.3373, 3.3793, 3.4062, 3.2099], + device='cuda:0'), covar=tensor([0.1131, 0.0685, 0.1214, 0.5752, 0.1024, 0.1235, 0.1653, 0.0937], + device='cuda:0'), in_proj_covar=tensor([0.0449, 0.0356, 0.0375, 0.0467, 0.0369, 0.0351, 0.0368, 0.0313], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 07:25:07,510 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69291.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:25:22,121 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69311.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:25:24,607 INFO [train.py:901] (0/4) Epoch 9, batch 4650, loss[loss=0.2256, simple_loss=0.2984, pruned_loss=0.07644, over 7653.00 frames. ], tot_loss[loss=0.2461, simple_loss=0.3188, pruned_loss=0.08667, over 1613341.15 frames. ], batch size: 19, lr: 8.49e-03, grad_scale: 8.0 +2023-02-06 07:25:38,790 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69336.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:25:40,085 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69338.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:25:42,427 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.16 vs. limit=5.0 +2023-02-06 07:25:42,602 INFO [optim.py:369] (0/4) 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 07:25:58,526 INFO [train.py:901] (0/4) Epoch 9, batch 4700, loss[loss=0.2393, simple_loss=0.3172, pruned_loss=0.08067, over 8339.00 frames. ], tot_loss[loss=0.2472, simple_loss=0.3194, pruned_loss=0.08746, over 1616485.24 frames. ], batch size: 25, lr: 8.49e-03, grad_scale: 8.0 +2023-02-06 07:26:14,060 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=69405.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:26:31,844 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69412.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:26:33,681 INFO [train.py:901] (0/4) Epoch 9, batch 4750, loss[loss=0.2237, simple_loss=0.3082, pruned_loss=0.06961, over 8323.00 frames. ], tot_loss[loss=0.2473, simple_loss=0.3196, pruned_loss=0.08749, over 1618111.70 frames. ], batch size: 25, lr: 8.48e-03, grad_scale: 8.0 +2023-02-06 07:26:35,885 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69418.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:26:47,322 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2054, 2.3997, 1.9393, 2.8135, 1.4242, 1.7711, 2.0348, 2.4420], + device='cuda:0'), covar=tensor([0.0675, 0.0809, 0.0995, 0.0428, 0.1161, 0.1270, 0.0960, 0.0711], + device='cuda:0'), in_proj_covar=tensor([0.0248, 0.0223, 0.0264, 0.0222, 0.0227, 0.0265, 0.0270, 0.0231], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 07:26:49,137 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-06 07:26:50,996 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-06 07:26:51,644 INFO [optim.py:369] (0/4) 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 07:26:53,203 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 9, batch 4800, loss[loss=0.2182, simple_loss=0.2977, pruned_loss=0.06936, over 8247.00 frames. ], tot_loss[loss=0.2473, simple_loss=0.3199, pruned_loss=0.08735, over 1620014.27 frames. ], batch size: 22, lr: 8.48e-03, grad_scale: 8.0 +2023-02-06 07:27:15,863 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.9020, 2.2393, 3.6729, 2.7096, 2.9045, 2.4039, 2.0585, 1.7934], + device='cuda:0'), covar=tensor([0.2995, 0.3989, 0.0961, 0.2303, 0.1949, 0.1830, 0.1441, 0.3902], + device='cuda:0'), in_proj_covar=tensor([0.0859, 0.0826, 0.0710, 0.0816, 0.0911, 0.0768, 0.0688, 0.0743], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 07:27:41,298 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-06 07:27:43,207 INFO [train.py:901] (0/4) Epoch 9, batch 4850, loss[loss=0.2255, simple_loss=0.3067, pruned_loss=0.07219, over 8239.00 frames. ], tot_loss[loss=0.2475, simple_loss=0.3198, pruned_loss=0.08757, over 1617880.31 frames. ], batch size: 24, lr: 8.48e-03, grad_scale: 8.0 +2023-02-06 07:27:46,744 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69520.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:27:49,409 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=69524.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:27:53,327 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=69530.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:28:00,627 INFO [optim.py:369] (0/4) 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 07:28:17,593 INFO [train.py:901] (0/4) Epoch 9, batch 4900, loss[loss=0.2758, simple_loss=0.3473, pruned_loss=0.1022, over 8568.00 frames. ], tot_loss[loss=0.2493, simple_loss=0.3214, pruned_loss=0.08859, over 1615565.21 frames. ], batch size: 31, lr: 8.48e-03, grad_scale: 8.0 +2023-02-06 07:28:33,114 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=69607.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 07:28:53,429 INFO [train.py:901] (0/4) Epoch 9, batch 4950, loss[loss=0.2485, simple_loss=0.334, pruned_loss=0.08153, over 8257.00 frames. ], tot_loss[loss=0.2489, simple_loss=0.3212, pruned_loss=0.08827, over 1612369.48 frames. ], batch size: 22, lr: 8.47e-03, grad_scale: 8.0 +2023-02-06 07:29:06,606 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=69635.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:29:09,347 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69639.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:29:10,456 INFO [optim.py:369] (0/4) 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 07:29:13,218 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69645.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:29:27,240 INFO [train.py:901] (0/4) Epoch 9, batch 5000, loss[loss=0.2587, simple_loss=0.3146, pruned_loss=0.1014, over 7815.00 frames. ], tot_loss[loss=0.2496, simple_loss=0.3219, pruned_loss=0.0887, over 1612315.59 frames. ], batch size: 20, lr: 8.47e-03, grad_scale: 8.0 +2023-02-06 07:29:39,108 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=69682.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:29:46,861 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69692.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 07:29:53,586 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69702.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:30:01,399 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 +2023-02-06 07:30:02,818 INFO [train.py:901] (0/4) Epoch 9, batch 5050, loss[loss=0.2617, simple_loss=0.3163, pruned_loss=0.1036, over 7704.00 frames. ], tot_loss[loss=0.2478, simple_loss=0.3199, pruned_loss=0.08784, over 1609497.51 frames. ], batch size: 18, lr: 8.47e-03, grad_scale: 8.0 +2023-02-06 07:30:07,754 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69722.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 07:30:12,761 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=69730.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:30:18,139 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-06 07:30:20,812 INFO [optim.py:369] (0/4) 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 07:30:26,990 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69750.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:30:29,703 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6184, 1.3431, 2.8414, 1.1711, 2.0022, 2.9855, 3.0953, 2.5114], + device='cuda:0'), covar=tensor([0.1110, 0.1555, 0.0433, 0.2201, 0.0876, 0.0367, 0.0522, 0.0800], + device='cuda:0'), in_proj_covar=tensor([0.0255, 0.0289, 0.0250, 0.0278, 0.0264, 0.0231, 0.0307, 0.0291], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 07:30:30,968 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=69756.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:30:36,707 INFO [train.py:901] (0/4) Epoch 9, batch 5100, loss[loss=0.2494, simple_loss=0.3196, pruned_loss=0.08961, over 7649.00 frames. ], tot_loss[loss=0.2487, simple_loss=0.3206, pruned_loss=0.08838, over 1610563.71 frames. ], batch size: 19, lr: 8.46e-03, grad_scale: 8.0 +2023-02-06 07:30:44,222 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69776.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:30:50,853 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0892, 1.6334, 1.5870, 1.2499, 0.9799, 1.3882, 1.5201, 1.5408], + device='cuda:0'), covar=tensor([0.0546, 0.1246, 0.1877, 0.1483, 0.0654, 0.1614, 0.0746, 0.0656], + device='cuda:0'), in_proj_covar=tensor([0.0106, 0.0160, 0.0199, 0.0163, 0.0110, 0.0168, 0.0122, 0.0140], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0006], + device='cuda:0') +2023-02-06 07:30:59,030 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69797.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:31:01,905 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69801.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:31:03,212 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7517, 3.9219, 2.4951, 2.6348, 2.7911, 2.0877, 2.5717, 2.8436], + device='cuda:0'), covar=tensor([0.1813, 0.0283, 0.0918, 0.0864, 0.0748, 0.1248, 0.1153, 0.1188], + device='cuda:0'), in_proj_covar=tensor([0.0352, 0.0241, 0.0315, 0.0301, 0.0310, 0.0324, 0.0345, 0.0314], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 07:31:11,529 INFO [train.py:901] (0/4) Epoch 9, batch 5150, loss[loss=0.2293, simple_loss=0.3076, pruned_loss=0.0755, over 8098.00 frames. ], tot_loss[loss=0.2466, simple_loss=0.3188, pruned_loss=0.0872, over 1609126.05 frames. ], batch size: 23, lr: 8.46e-03, grad_scale: 8.0 +2023-02-06 07:31:29,714 INFO [optim.py:369] (0/4) 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 07:31:32,788 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 9, batch 5200, loss[loss=0.2485, simple_loss=0.3251, pruned_loss=0.08593, over 8506.00 frames. ], tot_loss[loss=0.246, simple_loss=0.3187, pruned_loss=0.0867, over 1613278.37 frames. ], batch size: 26, lr: 8.46e-03, grad_scale: 8.0 +2023-02-06 07:31:50,875 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69871.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:32:02,077 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3178, 1.8740, 2.8721, 2.2078, 2.5458, 2.0695, 1.7147, 1.1876], + device='cuda:0'), covar=tensor([0.3208, 0.3367, 0.0903, 0.2463, 0.1650, 0.1931, 0.1544, 0.3659], + device='cuda:0'), in_proj_covar=tensor([0.0853, 0.0820, 0.0700, 0.0807, 0.0907, 0.0758, 0.0686, 0.0735], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 07:32:06,757 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69895.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:32:11,376 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69901.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:32:17,771 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 9, batch 5250, loss[loss=0.2405, simple_loss=0.3129, pruned_loss=0.08402, over 8027.00 frames. ], tot_loss[loss=0.2457, simple_loss=0.3181, pruned_loss=0.08661, over 1613703.93 frames. ], batch size: 22, lr: 8.45e-03, grad_scale: 8.0 +2023-02-06 07:32:23,748 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69920.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:32:27,836 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69926.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:32:38,261 INFO [optim.py:369] (0/4) 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] (0/4) 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] (0/4) Epoch 9, batch 5300, loss[loss=0.2175, simple_loss=0.3007, pruned_loss=0.06714, over 8126.00 frames. ], tot_loss[loss=0.2453, simple_loss=0.3179, pruned_loss=0.08637, over 1615139.46 frames. ], batch size: 22, lr: 8.45e-03, grad_scale: 8.0 +2023-02-06 07:33:04,976 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69978.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 07:33:08,299 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69983.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:33:19,761 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-70000.pt +2023-02-06 07:33:22,906 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=70003.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 07:33:24,830 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=70006.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:33:31,224 INFO [train.py:901] (0/4) Epoch 9, batch 5350, loss[loss=0.2451, simple_loss=0.313, pruned_loss=0.08862, over 7655.00 frames. ], tot_loss[loss=0.2472, simple_loss=0.319, pruned_loss=0.08773, over 1615328.07 frames. ], batch size: 19, lr: 8.45e-03, grad_scale: 8.0 +2023-02-06 07:33:42,004 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=70031.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:33:45,278 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=70036.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 07:33:48,395 INFO [optim.py:369] (0/4) 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 07:33:57,281 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=70053.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:34:05,194 INFO [train.py:901] (0/4) Epoch 9, batch 5400, loss[loss=0.24, simple_loss=0.3193, pruned_loss=0.08038, over 8025.00 frames. ], tot_loss[loss=0.2494, simple_loss=0.3209, pruned_loss=0.08895, over 1613697.47 frames. ], batch size: 22, lr: 8.45e-03, grad_scale: 8.0 +2023-02-06 07:34:14,803 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=70078.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:34:31,571 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 9, batch 5450, loss[loss=0.2105, simple_loss=0.2989, pruned_loss=0.06099, over 8243.00 frames. ], tot_loss[loss=0.249, simple_loss=0.3208, pruned_loss=0.0886, over 1614787.65 frames. ], batch size: 24, lr: 8.44e-03, grad_scale: 8.0 +2023-02-06 07:34:48,367 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=70126.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:34:49,042 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=70127.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:34:49,618 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=70128.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:34:58,813 INFO [optim.py:369] (0/4) 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 07:35:04,363 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-06 07:35:05,796 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70151.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 07:35:06,511 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=70152.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:35:15,858 INFO [train.py:901] (0/4) Epoch 9, batch 5500, loss[loss=0.2113, simple_loss=0.2832, pruned_loss=0.06972, over 7649.00 frames. ], tot_loss[loss=0.2473, simple_loss=0.3196, pruned_loss=0.08753, over 1614629.32 frames. ], batch size: 19, lr: 8.44e-03, grad_scale: 16.0 +2023-02-06 07:35:50,277 INFO [train.py:901] (0/4) Epoch 9, batch 5550, loss[loss=0.2448, simple_loss=0.3252, pruned_loss=0.08222, over 8319.00 frames. ], tot_loss[loss=0.2467, simple_loss=0.319, pruned_loss=0.08716, over 1615275.17 frames. ], batch size: 25, lr: 8.44e-03, grad_scale: 16.0 +2023-02-06 07:36:07,864 INFO [optim.py:369] (0/4) 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 07:36:24,758 INFO [train.py:901] (0/4) Epoch 9, batch 5600, loss[loss=0.2091, simple_loss=0.2992, pruned_loss=0.0595, over 8126.00 frames. ], tot_loss[loss=0.2455, simple_loss=0.3182, pruned_loss=0.08637, over 1615209.47 frames. ], batch size: 22, lr: 8.43e-03, grad_scale: 16.0 +2023-02-06 07:36:34,072 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=70278.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:36:35,514 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5955, 1.9772, 3.3676, 1.2315, 2.6318, 2.1134, 1.7937, 2.3531], + device='cuda:0'), covar=tensor([0.1762, 0.2276, 0.0882, 0.4000, 0.1412, 0.2594, 0.1681, 0.2062], + device='cuda:0'), in_proj_covar=tensor([0.0479, 0.0496, 0.0524, 0.0564, 0.0600, 0.0537, 0.0463, 0.0603], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 07:36:59,387 INFO [train.py:901] (0/4) Epoch 9, batch 5650, loss[loss=0.2078, simple_loss=0.2799, pruned_loss=0.06787, over 7940.00 frames. ], tot_loss[loss=0.2444, simple_loss=0.3173, pruned_loss=0.08572, over 1616678.90 frames. ], batch size: 20, lr: 8.43e-03, grad_scale: 8.0 +2023-02-06 07:37:08,096 WARNING [train.py:1067] (0/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-06 07:37:18,052 INFO [optim.py:369] (0/4) 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 07:37:32,964 INFO [train.py:901] (0/4) Epoch 9, batch 5700, loss[loss=0.2825, simple_loss=0.3536, pruned_loss=0.1057, over 8456.00 frames. ], tot_loss[loss=0.2471, simple_loss=0.3197, pruned_loss=0.08723, over 1618557.45 frames. ], batch size: 29, lr: 8.43e-03, grad_scale: 8.0 +2023-02-06 07:37:56,390 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.6830, 2.2256, 3.6840, 2.9268, 3.2759, 2.3733, 1.9176, 1.7458], + device='cuda:0'), covar=tensor([0.3173, 0.3716, 0.0992, 0.2105, 0.1753, 0.1824, 0.1432, 0.3929], + device='cuda:0'), in_proj_covar=tensor([0.0861, 0.0829, 0.0707, 0.0806, 0.0905, 0.0765, 0.0686, 0.0737], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 07:38:02,536 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.7127, 4.6524, 4.2256, 2.5945, 4.1006, 4.2404, 4.3595, 3.9333], + device='cuda:0'), covar=tensor([0.0798, 0.0477, 0.0804, 0.4594, 0.0822, 0.1081, 0.1124, 0.0869], + device='cuda:0'), in_proj_covar=tensor([0.0449, 0.0353, 0.0367, 0.0468, 0.0365, 0.0350, 0.0362, 0.0309], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 07:38:02,659 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=70407.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 07:38:08,014 INFO [train.py:901] (0/4) Epoch 9, batch 5750, loss[loss=0.2431, simple_loss=0.329, pruned_loss=0.07859, over 8249.00 frames. ], tot_loss[loss=0.2477, simple_loss=0.32, pruned_loss=0.08775, over 1614362.73 frames. ], batch size: 24, lr: 8.42e-03, grad_scale: 8.0 +2023-02-06 07:38:12,148 WARNING [train.py:1067] (0/4) 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] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=70432.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 07:38:27,529 INFO [optim.py:369] (0/4) 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 07:38:43,388 INFO [train.py:901] (0/4) Epoch 9, batch 5800, loss[loss=0.2322, simple_loss=0.3102, pruned_loss=0.0771, over 8122.00 frames. ], tot_loss[loss=0.2467, simple_loss=0.3192, pruned_loss=0.08705, over 1615703.40 frames. ], batch size: 22, lr: 8.42e-03, grad_scale: 8.0 +2023-02-06 07:38:48,302 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=70472.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:38:48,579 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.61 vs. limit=5.0 +2023-02-06 07:39:18,648 INFO [train.py:901] (0/4) Epoch 9, batch 5850, loss[loss=0.2158, simple_loss=0.2975, pruned_loss=0.06703, over 8321.00 frames. ], tot_loss[loss=0.2455, simple_loss=0.3182, pruned_loss=0.08641, over 1613712.27 frames. ], batch size: 25, lr: 8.42e-03, grad_scale: 8.0 +2023-02-06 07:39:37,364 INFO [optim.py:369] (0/4) 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] (0/4) attn_weights_entropy = tensor([2.2743, 1.8203, 2.8590, 2.2793, 2.4559, 2.1492, 1.6466, 1.1874], + device='cuda:0'), covar=tensor([0.3261, 0.3637, 0.0893, 0.1965, 0.1615, 0.1945, 0.1585, 0.3657], + device='cuda:0'), in_proj_covar=tensor([0.0853, 0.0819, 0.0703, 0.0805, 0.0900, 0.0761, 0.0683, 0.0733], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 07:39:53,321 INFO [train.py:901] (0/4) Epoch 9, batch 5900, loss[loss=0.246, simple_loss=0.321, pruned_loss=0.08553, over 8137.00 frames. ], tot_loss[loss=0.246, simple_loss=0.3186, pruned_loss=0.08674, over 1614958.20 frames. ], batch size: 22, lr: 8.42e-03, grad_scale: 8.0 +2023-02-06 07:40:08,006 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70587.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:40:27,232 INFO [train.py:901] (0/4) Epoch 9, batch 5950, loss[loss=0.258, simple_loss=0.3317, pruned_loss=0.09216, over 8513.00 frames. ], tot_loss[loss=0.2456, simple_loss=0.3183, pruned_loss=0.08647, over 1614794.88 frames. ], batch size: 29, lr: 8.41e-03, grad_scale: 8.0 +2023-02-06 07:40:32,699 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 07:41:02,078 INFO [train.py:901] (0/4) Epoch 9, batch 6000, loss[loss=0.2554, simple_loss=0.3294, pruned_loss=0.09068, over 8241.00 frames. ], tot_loss[loss=0.2438, simple_loss=0.3167, pruned_loss=0.08548, over 1613966.62 frames. ], batch size: 24, lr: 8.41e-03, grad_scale: 8.0 +2023-02-06 07:41:02,079 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 07:41:14,595 INFO [train.py:935] (0/4) Epoch 9, validation: loss=0.1952, simple_loss=0.2947, pruned_loss=0.0479, over 944034.00 frames. +2023-02-06 07:41:14,596 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-06 07:41:49,714 INFO [train.py:901] (0/4) Epoch 9, batch 6050, loss[loss=0.2735, simple_loss=0.3479, pruned_loss=0.09959, over 7981.00 frames. ], tot_loss[loss=0.2459, simple_loss=0.3185, pruned_loss=0.08665, over 1612574.10 frames. ], batch size: 21, lr: 8.41e-03, grad_scale: 8.0 +2023-02-06 07:42:03,382 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6006, 2.8925, 1.7822, 2.2727, 2.2945, 1.4892, 2.0659, 2.1768], + device='cuda:0'), covar=tensor([0.1187, 0.0272, 0.0929, 0.0615, 0.0583, 0.1281, 0.0912, 0.0747], + device='cuda:0'), in_proj_covar=tensor([0.0344, 0.0234, 0.0307, 0.0294, 0.0304, 0.0314, 0.0335, 0.0307], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 07:42:04,738 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70737.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:42:07,995 INFO [optim.py:369] (0/4) 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 07:42:24,368 INFO [train.py:901] (0/4) Epoch 9, batch 6100, loss[loss=0.2633, simple_loss=0.3458, pruned_loss=0.0904, over 8560.00 frames. ], tot_loss[loss=0.2452, simple_loss=0.318, pruned_loss=0.08625, over 1607966.83 frames. ], batch size: 39, lr: 8.40e-03, grad_scale: 8.0 +2023-02-06 07:42:42,094 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-06 07:42:53,129 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.7305, 3.0174, 2.4159, 4.1585, 1.8070, 2.0291, 2.3235, 3.4290], + device='cuda:0'), covar=tensor([0.0757, 0.0884, 0.1040, 0.0217, 0.1241, 0.1549, 0.1401, 0.0768], + device='cuda:0'), in_proj_covar=tensor([0.0246, 0.0220, 0.0262, 0.0219, 0.0223, 0.0261, 0.0267, 0.0227], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 07:43:00,373 INFO [train.py:901] (0/4) Epoch 9, batch 6150, loss[loss=0.2663, simple_loss=0.347, pruned_loss=0.09282, over 8035.00 frames. ], tot_loss[loss=0.2439, simple_loss=0.3174, pruned_loss=0.08525, over 1614287.06 frames. ], batch size: 22, lr: 8.40e-03, grad_scale: 8.0 +2023-02-06 07:43:18,323 INFO [optim.py:369] (0/4) 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 07:43:19,249 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=70843.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:43:22,053 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7195, 2.2443, 4.5818, 1.2739, 2.9577, 2.3207, 1.6732, 2.7661], + device='cuda:0'), covar=tensor([0.1658, 0.2163, 0.0605, 0.3759, 0.1643, 0.2535, 0.1752, 0.2317], + device='cuda:0'), in_proj_covar=tensor([0.0488, 0.0503, 0.0533, 0.0574, 0.0616, 0.0548, 0.0469, 0.0615], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:0') +2023-02-06 07:43:33,938 INFO [train.py:901] (0/4) Epoch 9, batch 6200, loss[loss=0.2557, simple_loss=0.3315, pruned_loss=0.08993, over 8329.00 frames. ], tot_loss[loss=0.2426, simple_loss=0.3164, pruned_loss=0.08443, over 1612482.23 frames. ], batch size: 26, lr: 8.40e-03, grad_scale: 8.0 +2023-02-06 07:43:36,216 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=70868.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:44:08,529 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5367, 1.5385, 1.7532, 1.4377, 0.9638, 1.7615, 0.0752, 1.2928], + device='cuda:0'), covar=tensor([0.3122, 0.2051, 0.0606, 0.1585, 0.5252, 0.0703, 0.3695, 0.1691], + device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0157, 0.0090, 0.0207, 0.0242, 0.0095, 0.0158, 0.0155], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 07:44:09,710 INFO [train.py:901] (0/4) Epoch 9, batch 6250, loss[loss=0.2359, simple_loss=0.3136, pruned_loss=0.07914, over 7812.00 frames. ], tot_loss[loss=0.244, simple_loss=0.317, pruned_loss=0.0855, over 1612006.98 frames. ], batch size: 20, lr: 8.40e-03, grad_scale: 8.0 +2023-02-06 07:44:28,454 INFO [optim.py:369] (0/4) 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 07:44:43,938 INFO [train.py:901] (0/4) Epoch 9, batch 6300, loss[loss=0.2526, simple_loss=0.3277, pruned_loss=0.08877, over 8236.00 frames. ], tot_loss[loss=0.2424, simple_loss=0.3155, pruned_loss=0.08467, over 1614115.48 frames. ], batch size: 22, lr: 8.39e-03, grad_scale: 8.0 +2023-02-06 07:44:47,606 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5414, 1.9278, 3.2815, 1.3347, 2.3611, 2.0031, 1.6755, 2.1660], + device='cuda:0'), covar=tensor([0.1670, 0.2138, 0.0774, 0.3616, 0.1614, 0.2672, 0.1661, 0.2298], + device='cuda:0'), in_proj_covar=tensor([0.0488, 0.0504, 0.0533, 0.0576, 0.0614, 0.0550, 0.0468, 0.0611], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 07:45:03,875 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=70993.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:45:16,112 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 9, batch 6350, loss[loss=0.2224, simple_loss=0.2983, pruned_loss=0.07329, over 7796.00 frames. ], tot_loss[loss=0.2418, simple_loss=0.3151, pruned_loss=0.08429, over 1614493.59 frames. ], batch size: 19, lr: 8.39e-03, grad_scale: 8.0 +2023-02-06 07:45:21,635 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=71018.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:45:23,674 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5022, 1.9770, 3.3452, 1.2913, 2.4474, 1.9298, 1.6450, 2.2803], + device='cuda:0'), covar=tensor([0.1737, 0.2141, 0.0626, 0.3774, 0.1500, 0.2677, 0.1633, 0.2038], + device='cuda:0'), in_proj_covar=tensor([0.0486, 0.0501, 0.0528, 0.0573, 0.0612, 0.0547, 0.0465, 0.0609], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 07:45:38,849 INFO [optim.py:369] (0/4) 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] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=71048.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:45:54,197 INFO [train.py:901] (0/4) Epoch 9, batch 6400, loss[loss=0.2686, simple_loss=0.3501, pruned_loss=0.09354, over 8102.00 frames. ], tot_loss[loss=0.2426, simple_loss=0.3161, pruned_loss=0.08454, over 1616068.69 frames. ], batch size: 23, lr: 8.39e-03, grad_scale: 8.0 +2023-02-06 07:46:28,849 INFO [train.py:901] (0/4) Epoch 9, batch 6450, loss[loss=0.22, simple_loss=0.3061, pruned_loss=0.06698, over 8369.00 frames. ], tot_loss[loss=0.2422, simple_loss=0.3154, pruned_loss=0.08452, over 1615052.82 frames. ], batch size: 24, lr: 8.38e-03, grad_scale: 4.0 +2023-02-06 07:46:48,393 INFO [optim.py:369] (0/4) 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 07:47:03,607 INFO [train.py:901] (0/4) Epoch 9, batch 6500, loss[loss=0.2043, simple_loss=0.2774, pruned_loss=0.06557, over 7227.00 frames. ], tot_loss[loss=0.2429, simple_loss=0.3158, pruned_loss=0.08503, over 1615597.59 frames. ], batch size: 16, lr: 8.38e-03, grad_scale: 4.0 +2023-02-06 07:47:03,824 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4853, 1.4820, 1.7420, 1.4427, 0.9241, 1.7519, 0.0674, 1.1975], + device='cuda:0'), covar=tensor([0.2662, 0.2096, 0.0655, 0.1917, 0.5267, 0.0782, 0.4036, 0.1825], + device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0158, 0.0094, 0.0208, 0.0242, 0.0096, 0.0158, 0.0156], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 07:47:37,714 INFO [train.py:901] (0/4) Epoch 9, batch 6550, loss[loss=0.1999, simple_loss=0.2703, pruned_loss=0.06476, over 7546.00 frames. ], tot_loss[loss=0.242, simple_loss=0.3146, pruned_loss=0.08466, over 1613653.06 frames. ], batch size: 18, lr: 8.38e-03, grad_scale: 4.0 +2023-02-06 07:47:40,259 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 +2023-02-06 07:47:50,022 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-06 07:47:52,260 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5426, 1.9434, 1.9860, 1.0032, 2.2309, 1.3500, 0.5332, 1.7239], + device='cuda:0'), covar=tensor([0.0375, 0.0176, 0.0151, 0.0332, 0.0185, 0.0502, 0.0477, 0.0163], + device='cuda:0'), in_proj_covar=tensor([0.0375, 0.0302, 0.0246, 0.0358, 0.0287, 0.0448, 0.0340, 0.0324], + device='cuda:0'), out_proj_covar=tensor([1.1096e-04, 8.6896e-05, 7.1166e-05, 1.0374e-04, 8.4438e-05, 1.4164e-04, + 1.0073e-04, 9.5381e-05], device='cuda:0') +2023-02-06 07:47:58,021 INFO [optim.py:369] (0/4) 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 07:48:10,529 WARNING [train.py:1067] (0/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-06 07:48:13,163 INFO [train.py:901] (0/4) Epoch 9, batch 6600, loss[loss=0.2664, simple_loss=0.3487, pruned_loss=0.09208, over 8358.00 frames. ], tot_loss[loss=0.2428, simple_loss=0.3153, pruned_loss=0.0852, over 1612155.62 frames. ], batch size: 24, lr: 8.37e-03, grad_scale: 4.0 +2023-02-06 07:48:47,048 INFO [train.py:901] (0/4) Epoch 9, batch 6650, loss[loss=0.2946, simple_loss=0.3518, pruned_loss=0.1186, over 7055.00 frames. ], tot_loss[loss=0.2447, simple_loss=0.3173, pruned_loss=0.08602, over 1616522.34 frames. ], batch size: 74, lr: 8.37e-03, grad_scale: 4.0 +2023-02-06 07:48:51,522 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 +2023-02-06 07:48:55,997 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6251, 3.7215, 2.2576, 2.3489, 2.6466, 1.9517, 2.2921, 2.8301], + device='cuda:0'), covar=tensor([0.1676, 0.0310, 0.0990, 0.0841, 0.0714, 0.1302, 0.1154, 0.0996], + device='cuda:0'), in_proj_covar=tensor([0.0340, 0.0235, 0.0307, 0.0295, 0.0302, 0.0315, 0.0334, 0.0306], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 07:49:05,724 INFO [optim.py:369] (0/4) 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 07:49:11,620 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.62 vs. limit=5.0 +2023-02-06 07:49:13,927 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=71354.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:49:21,682 INFO [train.py:901] (0/4) Epoch 9, batch 6700, loss[loss=0.2468, simple_loss=0.3297, pruned_loss=0.0819, over 8331.00 frames. ], tot_loss[loss=0.2453, simple_loss=0.3173, pruned_loss=0.08661, over 1613915.69 frames. ], batch size: 25, lr: 8.37e-03, grad_scale: 4.0 +2023-02-06 07:49:41,045 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=71392.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:49:56,536 INFO [train.py:901] (0/4) Epoch 9, batch 6750, loss[loss=0.2643, simple_loss=0.3428, pruned_loss=0.09296, over 8289.00 frames. ], tot_loss[loss=0.2454, simple_loss=0.3178, pruned_loss=0.08648, over 1616246.31 frames. ], batch size: 23, lr: 8.37e-03, grad_scale: 4.0 +2023-02-06 07:50:15,374 INFO [optim.py:369] (0/4) 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 07:50:23,400 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-06 07:50:30,408 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3838, 1.5460, 2.3067, 1.2301, 1.6486, 1.6404, 1.4468, 1.5461], + device='cuda:0'), covar=tensor([0.1733, 0.2015, 0.0720, 0.3469, 0.1372, 0.2851, 0.1732, 0.1762], + device='cuda:0'), in_proj_covar=tensor([0.0487, 0.0500, 0.0529, 0.0575, 0.0611, 0.0548, 0.0469, 0.0612], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 07:50:30,842 INFO [train.py:901] (0/4) Epoch 9, batch 6800, loss[loss=0.2554, simple_loss=0.3156, pruned_loss=0.09756, over 7975.00 frames. ], tot_loss[loss=0.2453, simple_loss=0.3181, pruned_loss=0.08622, over 1618886.61 frames. ], batch size: 21, lr: 8.36e-03, grad_scale: 8.0 +2023-02-06 07:50:33,613 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=71469.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:51:01,251 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=71507.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:51:06,469 INFO [train.py:901] (0/4) Epoch 9, batch 6850, loss[loss=0.2673, simple_loss=0.3221, pruned_loss=0.1063, over 5979.00 frames. ], tot_loss[loss=0.246, simple_loss=0.3184, pruned_loss=0.08678, over 1612713.48 frames. ], batch size: 13, lr: 8.36e-03, grad_scale: 8.0 +2023-02-06 07:51:14,498 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-06 07:51:25,245 INFO [optim.py:369] (0/4) 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 07:51:40,045 INFO [train.py:901] (0/4) Epoch 9, batch 6900, loss[loss=0.3051, simple_loss=0.3704, pruned_loss=0.1199, over 8284.00 frames. ], tot_loss[loss=0.2447, simple_loss=0.3173, pruned_loss=0.086, over 1610925.36 frames. ], batch size: 23, lr: 8.36e-03, grad_scale: 8.0 +2023-02-06 07:51:48,539 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4563, 1.8124, 4.3527, 1.6589, 2.4184, 4.8397, 4.9014, 4.0268], + device='cuda:0'), covar=tensor([0.1056, 0.1519, 0.0308, 0.2093, 0.1058, 0.0261, 0.0365, 0.0801], + device='cuda:0'), in_proj_covar=tensor([0.0253, 0.0282, 0.0249, 0.0278, 0.0261, 0.0226, 0.0305, 0.0287], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 07:52:15,385 INFO [train.py:901] (0/4) Epoch 9, batch 6950, loss[loss=0.2202, simple_loss=0.3028, pruned_loss=0.06883, over 7973.00 frames. ], tot_loss[loss=0.2443, simple_loss=0.3173, pruned_loss=0.08564, over 1612682.23 frames. ], batch size: 21, lr: 8.35e-03, grad_scale: 8.0 +2023-02-06 07:52:23,477 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-06 07:52:29,643 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=71634.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:52:35,516 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 9, batch 7000, loss[loss=0.2299, simple_loss=0.3112, pruned_loss=0.0743, over 8244.00 frames. ], tot_loss[loss=0.2442, simple_loss=0.3175, pruned_loss=0.08549, over 1616435.26 frames. ], batch size: 24, lr: 8.35e-03, grad_scale: 8.0 +2023-02-06 07:53:19,913 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 +2023-02-06 07:53:21,190 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.81 vs. limit=2.0 +2023-02-06 07:53:24,881 INFO [train.py:901] (0/4) Epoch 9, batch 7050, loss[loss=0.202, simple_loss=0.2841, pruned_loss=0.05993, over 8454.00 frames. ], tot_loss[loss=0.2436, simple_loss=0.3167, pruned_loss=0.08522, over 1610703.65 frames. ], batch size: 25, lr: 8.35e-03, grad_scale: 8.0 +2023-02-06 07:53:32,430 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 07:53:48,776 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.6024, 1.6086, 5.7420, 2.0466, 5.0034, 4.7661, 5.3253, 5.0939], + device='cuda:0'), covar=tensor([0.0482, 0.4306, 0.0362, 0.3315, 0.1066, 0.0813, 0.0451, 0.0521], + device='cuda:0'), in_proj_covar=tensor([0.0447, 0.0552, 0.0553, 0.0513, 0.0582, 0.0495, 0.0485, 0.0545], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 07:53:50,707 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=71750.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:53:59,278 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=71763.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:54:00,447 INFO [train.py:901] (0/4) Epoch 9, batch 7100, loss[loss=0.2577, simple_loss=0.3235, pruned_loss=0.096, over 8249.00 frames. ], tot_loss[loss=0.2448, simple_loss=0.3177, pruned_loss=0.08588, over 1613137.30 frames. ], batch size: 22, lr: 8.35e-03, grad_scale: 8.0 +2023-02-06 07:54:08,147 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-02-06 07:54:16,148 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=71788.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:54:34,522 INFO [train.py:901] (0/4) Epoch 9, batch 7150, loss[loss=0.25, simple_loss=0.3258, pruned_loss=0.08712, over 8581.00 frames. ], tot_loss[loss=0.2434, simple_loss=0.3168, pruned_loss=0.08502, over 1616801.17 frames. ], batch size: 31, lr: 8.34e-03, grad_scale: 8.0 +2023-02-06 07:54:54,742 INFO [optim.py:369] (0/4) 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 07:55:10,753 INFO [train.py:901] (0/4) Epoch 9, batch 7200, loss[loss=0.2455, simple_loss=0.305, pruned_loss=0.093, over 7420.00 frames. ], tot_loss[loss=0.244, simple_loss=0.3168, pruned_loss=0.08557, over 1617453.73 frames. ], batch size: 17, lr: 8.34e-03, grad_scale: 8.0 +2023-02-06 07:55:37,711 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 +2023-02-06 07:55:40,058 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.1528, 2.9966, 3.4519, 2.3758, 1.8314, 3.6014, 0.8119, 2.1562], + device='cuda:0'), covar=tensor([0.1663, 0.1340, 0.0469, 0.2308, 0.4445, 0.0290, 0.4069, 0.2319], + device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0157, 0.0092, 0.0205, 0.0244, 0.0095, 0.0154, 0.0155], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0001, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 07:55:41,108 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.24 vs. limit=5.0 +2023-02-06 07:55:43,960 INFO [train.py:901] (0/4) Epoch 9, batch 7250, loss[loss=0.2405, simple_loss=0.3243, pruned_loss=0.0783, over 8505.00 frames. ], tot_loss[loss=0.2458, simple_loss=0.3184, pruned_loss=0.08655, over 1619719.01 frames. ], batch size: 26, lr: 8.34e-03, grad_scale: 8.0 +2023-02-06 07:55:58,950 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) Epoch 9, batch 7300, loss[loss=0.2281, simple_loss=0.2977, pruned_loss=0.07926, over 7425.00 frames. ], tot_loss[loss=0.2459, simple_loss=0.3184, pruned_loss=0.08667, over 1619153.76 frames. ], batch size: 17, lr: 8.33e-03, grad_scale: 8.0 +2023-02-06 07:56:28,819 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=71978.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:56:29,624 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2907, 2.6748, 2.1902, 4.0109, 1.8795, 1.9164, 2.1447, 3.0277], + device='cuda:0'), covar=tensor([0.0777, 0.1027, 0.1034, 0.0206, 0.1112, 0.1411, 0.1258, 0.0874], + device='cuda:0'), in_proj_covar=tensor([0.0241, 0.0218, 0.0258, 0.0215, 0.0219, 0.0257, 0.0260, 0.0224], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 07:56:33,716 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3635, 2.0296, 3.3938, 1.2217, 2.4996, 1.7967, 1.5411, 2.2940], + device='cuda:0'), covar=tensor([0.1741, 0.2041, 0.0660, 0.3753, 0.1421, 0.2766, 0.1765, 0.2062], + device='cuda:0'), in_proj_covar=tensor([0.0480, 0.0493, 0.0527, 0.0569, 0.0605, 0.0541, 0.0462, 0.0606], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 07:56:43,618 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-72000.pt +2023-02-06 07:56:54,481 INFO [train.py:901] (0/4) Epoch 9, batch 7350, loss[loss=0.2016, simple_loss=0.279, pruned_loss=0.0621, over 7810.00 frames. ], tot_loss[loss=0.2458, simple_loss=0.3186, pruned_loss=0.08647, over 1621181.52 frames. ], batch size: 20, lr: 8.33e-03, grad_scale: 8.0 +2023-02-06 07:57:02,499 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-06 07:57:04,268 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.80 vs. limit=5.0 +2023-02-06 07:57:13,532 INFO [optim.py:369] (0/4) 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 07:57:22,475 WARNING [train.py:1067] (0/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-06 07:57:30,037 INFO [train.py:901] (0/4) Epoch 9, batch 7400, loss[loss=0.212, simple_loss=0.2763, pruned_loss=0.07387, over 7820.00 frames. ], tot_loss[loss=0.2452, simple_loss=0.3181, pruned_loss=0.08617, over 1621151.78 frames. ], batch size: 19, lr: 8.33e-03, grad_scale: 8.0 +2023-02-06 07:57:40,384 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6149, 2.3329, 4.6426, 1.4391, 3.0009, 2.1066, 1.7380, 2.8183], + device='cuda:0'), covar=tensor([0.1617, 0.2007, 0.0520, 0.3625, 0.1504, 0.2659, 0.1677, 0.2142], + device='cuda:0'), in_proj_covar=tensor([0.0481, 0.0494, 0.0530, 0.0569, 0.0608, 0.0539, 0.0463, 0.0609], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 07:57:50,406 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72093.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:58:03,889 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 9, batch 7450, loss[loss=0.2271, simple_loss=0.2887, pruned_loss=0.08274, over 7932.00 frames. ], tot_loss[loss=0.2455, simple_loss=0.3189, pruned_loss=0.08608, over 1622838.41 frames. ], batch size: 20, lr: 8.33e-03, grad_scale: 8.0 +2023-02-06 07:58:23,972 INFO [optim.py:369] (0/4) 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 07:58:26,821 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=72147.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 07:58:38,700 INFO [train.py:901] (0/4) Epoch 9, batch 7500, loss[loss=0.2336, simple_loss=0.3052, pruned_loss=0.081, over 8092.00 frames. ], tot_loss[loss=0.2452, simple_loss=0.3182, pruned_loss=0.08605, over 1620170.80 frames. ], batch size: 21, lr: 8.32e-03, grad_scale: 8.0 +2023-02-06 07:59:09,166 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1661, 2.6736, 3.1297, 1.1334, 3.1942, 1.9941, 1.5567, 2.0761], + device='cuda:0'), covar=tensor([0.0602, 0.0239, 0.0255, 0.0523, 0.0334, 0.0513, 0.0582, 0.0309], + device='cuda:0'), in_proj_covar=tensor([0.0381, 0.0311, 0.0257, 0.0371, 0.0298, 0.0456, 0.0349, 0.0333], + device='cuda:0'), out_proj_covar=tensor([1.1208e-04, 8.9315e-05, 7.4195e-05, 1.0741e-04, 8.7533e-05, 1.4361e-04, + 1.0318e-04, 9.7734e-05], device='cuda:0') +2023-02-06 07:59:15,018 INFO [train.py:901] (0/4) Epoch 9, batch 7550, loss[loss=0.239, simple_loss=0.3204, pruned_loss=0.0788, over 8326.00 frames. ], tot_loss[loss=0.2458, simple_loss=0.3183, pruned_loss=0.08666, over 1621467.65 frames. ], batch size: 25, lr: 8.32e-03, grad_scale: 8.0 +2023-02-06 07:59:33,687 INFO [optim.py:369] (0/4) 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 07:59:48,148 INFO [train.py:901] (0/4) Epoch 9, batch 7600, loss[loss=0.2761, simple_loss=0.3429, pruned_loss=0.1046, over 8457.00 frames. ], tot_loss[loss=0.2474, simple_loss=0.3193, pruned_loss=0.08774, over 1617537.14 frames. ], batch size: 27, lr: 8.32e-03, grad_scale: 8.0 +2023-02-06 07:59:58,686 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=72281.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:00:22,658 INFO [train.py:901] (0/4) Epoch 9, batch 7650, loss[loss=0.2655, simple_loss=0.321, pruned_loss=0.105, over 7939.00 frames. ], tot_loss[loss=0.2453, simple_loss=0.3178, pruned_loss=0.08638, over 1616534.80 frames. ], batch size: 20, lr: 8.31e-03, grad_scale: 8.0 +2023-02-06 08:00:42,776 INFO [optim.py:369] (0/4) 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 08:00:43,581 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=72344.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:00:47,004 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=72349.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:00:53,771 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3595, 2.6877, 1.8655, 2.2245, 2.1199, 1.4903, 1.9292, 2.0076], + device='cuda:0'), covar=tensor([0.1287, 0.0311, 0.0868, 0.0522, 0.0593, 0.1266, 0.0842, 0.0813], + device='cuda:0'), in_proj_covar=tensor([0.0342, 0.0237, 0.0312, 0.0298, 0.0303, 0.0322, 0.0336, 0.0308], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 08:00:57,577 INFO [train.py:901] (0/4) Epoch 9, batch 7700, loss[loss=0.208, simple_loss=0.2948, pruned_loss=0.06061, over 8326.00 frames. ], tot_loss[loss=0.2451, simple_loss=0.3175, pruned_loss=0.08638, over 1612416.20 frames. ], batch size: 25, lr: 8.31e-03, grad_scale: 8.0 +2023-02-06 08:01:03,906 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72374.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:01:09,788 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-06 08:01:18,720 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72396.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:01:22,917 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3041, 1.5307, 1.5171, 1.3603, 0.9247, 1.3781, 1.8296, 1.4890], + device='cuda:0'), covar=tensor([0.0502, 0.1238, 0.1872, 0.1455, 0.0615, 0.1573, 0.0685, 0.0629], + device='cuda:0'), in_proj_covar=tensor([0.0105, 0.0157, 0.0198, 0.0162, 0.0108, 0.0168, 0.0119, 0.0141], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0006, 0.0006], + device='cuda:0') +2023-02-06 08:01:32,219 INFO [train.py:901] (0/4) Epoch 9, batch 7750, loss[loss=0.2488, simple_loss=0.3189, pruned_loss=0.08935, over 7808.00 frames. ], tot_loss[loss=0.2452, simple_loss=0.318, pruned_loss=0.08626, over 1610801.46 frames. ], batch size: 20, lr: 8.31e-03, grad_scale: 8.0 +2023-02-06 08:01:53,040 INFO [optim.py:369] (0/4) 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 08:01:57,608 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3529, 1.7971, 2.7240, 1.1300, 1.9736, 1.6357, 1.5331, 1.6791], + device='cuda:0'), covar=tensor([0.1860, 0.2284, 0.0823, 0.4283, 0.1672, 0.3120, 0.1900, 0.2369], + device='cuda:0'), in_proj_covar=tensor([0.0478, 0.0495, 0.0528, 0.0571, 0.0607, 0.0537, 0.0463, 0.0608], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 08:02:08,200 INFO [train.py:901] (0/4) Epoch 9, batch 7800, loss[loss=0.2125, simple_loss=0.2968, pruned_loss=0.06408, over 8248.00 frames. ], tot_loss[loss=0.244, simple_loss=0.3166, pruned_loss=0.08569, over 1606766.93 frames. ], batch size: 24, lr: 8.31e-03, grad_scale: 8.0 +2023-02-06 08:02:10,518 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.6719, 2.0984, 3.2146, 2.4951, 2.8416, 2.3157, 1.9422, 1.5821], + device='cuda:0'), covar=tensor([0.2818, 0.3398, 0.0896, 0.2032, 0.1520, 0.1623, 0.1334, 0.3397], + device='cuda:0'), in_proj_covar=tensor([0.0851, 0.0818, 0.0692, 0.0805, 0.0892, 0.0754, 0.0681, 0.0736], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 08:02:25,495 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=72491.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:02:41,269 INFO [train.py:901] (0/4) Epoch 9, batch 7850, loss[loss=0.2454, simple_loss=0.3281, pruned_loss=0.08133, over 8471.00 frames. ], tot_loss[loss=0.2429, simple_loss=0.3159, pruned_loss=0.08495, over 1604306.44 frames. ], batch size: 25, lr: 8.30e-03, grad_scale: 8.0 +2023-02-06 08:02:59,530 INFO [optim.py:369] (0/4) 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 08:03:14,032 INFO [train.py:901] (0/4) Epoch 9, batch 7900, loss[loss=0.2389, simple_loss=0.294, pruned_loss=0.09185, over 7695.00 frames. ], tot_loss[loss=0.244, simple_loss=0.317, pruned_loss=0.08547, over 1609492.81 frames. ], batch size: 18, lr: 8.30e-03, grad_scale: 8.0 +2023-02-06 08:03:38,853 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0710, 3.7311, 2.1716, 2.9406, 2.6744, 1.5722, 2.6203, 2.8944], + device='cuda:0'), covar=tensor([0.1547, 0.0262, 0.1124, 0.0620, 0.0750, 0.1582, 0.1018, 0.1062], + device='cuda:0'), in_proj_covar=tensor([0.0341, 0.0233, 0.0309, 0.0294, 0.0299, 0.0318, 0.0332, 0.0308], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 08:03:41,420 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 9, batch 7950, loss[loss=0.2573, simple_loss=0.3292, pruned_loss=0.09265, over 8649.00 frames. ], tot_loss[loss=0.2457, simple_loss=0.3188, pruned_loss=0.08624, over 1613596.41 frames. ], batch size: 39, lr: 8.30e-03, grad_scale: 8.0 +2023-02-06 08:04:05,356 INFO [optim.py:369] (0/4) 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] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=72652.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:04:19,766 INFO [train.py:901] (0/4) Epoch 9, batch 8000, loss[loss=0.1991, simple_loss=0.2709, pruned_loss=0.06367, over 7789.00 frames. ], tot_loss[loss=0.246, simple_loss=0.3191, pruned_loss=0.08649, over 1608958.08 frames. ], batch size: 19, lr: 8.29e-03, grad_scale: 8.0 +2023-02-06 08:04:27,787 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72677.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:04:28,664 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 +2023-02-06 08:04:34,984 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 9, batch 8050, loss[loss=0.2209, simple_loss=0.2872, pruned_loss=0.07728, over 7553.00 frames. ], tot_loss[loss=0.2436, simple_loss=0.3156, pruned_loss=0.08579, over 1586000.70 frames. ], batch size: 18, lr: 8.29e-03, grad_scale: 8.0 +2023-02-06 08:05:11,531 INFO [optim.py:369] (0/4) 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 08:05:15,393 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/epoch-9.pt +2023-02-06 08:05:27,546 WARNING [train.py:1067] (0/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-06 08:05:31,306 INFO [train.py:901] (0/4) Epoch 10, batch 0, loss[loss=0.267, simple_loss=0.3297, pruned_loss=0.1021, over 8114.00 frames. ], tot_loss[loss=0.267, simple_loss=0.3297, pruned_loss=0.1021, over 8114.00 frames. ], batch size: 23, lr: 7.88e-03, grad_scale: 8.0 +2023-02-06 08:05:31,306 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 08:05:43,263 INFO [train.py:935] (0/4) Epoch 10, validation: loss=0.1954, simple_loss=0.295, pruned_loss=0.0479, over 944034.00 frames. +2023-02-06 08:05:43,264 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-06 08:05:57,152 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-06 08:06:17,969 INFO [train.py:901] (0/4) Epoch 10, batch 50, loss[loss=0.2856, simple_loss=0.3468, pruned_loss=0.1122, over 8148.00 frames. ], tot_loss[loss=0.2467, simple_loss=0.3185, pruned_loss=0.0874, over 363884.40 frames. ], batch size: 22, lr: 7.88e-03, grad_scale: 8.0 +2023-02-06 08:06:21,766 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72803.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:06:31,243 WARNING [train.py:1067] (0/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-06 08:06:42,350 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2497, 1.4466, 1.5906, 1.4193, 1.0854, 1.4058, 1.7389, 1.4886], + device='cuda:0'), covar=tensor([0.0517, 0.1274, 0.1762, 0.1444, 0.0601, 0.1537, 0.0710, 0.0644], + device='cuda:0'), in_proj_covar=tensor([0.0106, 0.0157, 0.0197, 0.0162, 0.0107, 0.0168, 0.0121, 0.0140], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0006, 0.0006], + device='cuda:0') +2023-02-06 08:06:49,409 INFO [optim.py:369] (0/4) 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 08:06:52,306 WARNING [train.py:1067] (0/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-06 08:06:52,968 INFO [train.py:901] (0/4) Epoch 10, batch 100, loss[loss=0.2959, simple_loss=0.3595, pruned_loss=0.1162, over 8547.00 frames. ], tot_loss[loss=0.2445, simple_loss=0.3175, pruned_loss=0.08574, over 642401.97 frames. ], batch size: 31, lr: 7.88e-03, grad_scale: 8.0 +2023-02-06 08:07:03,832 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=72862.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:07:22,329 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72887.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:07:22,976 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3217, 1.3423, 1.6069, 1.3879, 1.0816, 1.3051, 1.7949, 1.6326], + device='cuda:0'), covar=tensor([0.0499, 0.1342, 0.1781, 0.1455, 0.0596, 0.1623, 0.0683, 0.0611], + device='cuda:0'), in_proj_covar=tensor([0.0105, 0.0156, 0.0196, 0.0161, 0.0107, 0.0167, 0.0120, 0.0139], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0006, 0.0006], + device='cuda:0') +2023-02-06 08:07:30,284 INFO [train.py:901] (0/4) Epoch 10, batch 150, loss[loss=0.2551, simple_loss=0.3259, pruned_loss=0.09217, over 8509.00 frames. ], tot_loss[loss=0.2459, simple_loss=0.3189, pruned_loss=0.08648, over 862262.50 frames. ], batch size: 29, lr: 7.87e-03, grad_scale: 8.0 +2023-02-06 08:07:33,248 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4015, 1.9407, 3.2353, 1.1513, 2.2423, 1.8160, 1.5442, 2.1123], + device='cuda:0'), covar=tensor([0.1832, 0.2040, 0.0674, 0.4154, 0.1694, 0.2903, 0.1866, 0.2360], + device='cuda:0'), in_proj_covar=tensor([0.0485, 0.0498, 0.0540, 0.0575, 0.0615, 0.0544, 0.0470, 0.0616], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], + device='cuda:0') +2023-02-06 08:07:40,087 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=72912.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:08:01,162 INFO [optim.py:369] (0/4) 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 08:08:04,566 INFO [train.py:901] (0/4) Epoch 10, batch 200, loss[loss=0.2025, simple_loss=0.2916, pruned_loss=0.05666, over 7935.00 frames. ], tot_loss[loss=0.2439, simple_loss=0.318, pruned_loss=0.08487, over 1028845.18 frames. ], batch size: 20, lr: 7.87e-03, grad_scale: 8.0 +2023-02-06 08:08:10,995 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.5799, 3.0874, 2.7202, 4.3080, 1.9346, 1.9871, 2.3374, 3.3867], + device='cuda:0'), covar=tensor([0.0800, 0.0843, 0.0908, 0.0208, 0.1151, 0.1521, 0.1233, 0.0758], + device='cuda:0'), in_proj_covar=tensor([0.0243, 0.0220, 0.0261, 0.0218, 0.0223, 0.0257, 0.0263, 0.0225], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 08:08:29,449 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=72982.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 08:08:41,028 INFO [train.py:901] (0/4) Epoch 10, batch 250, loss[loss=0.2711, simple_loss=0.3403, pruned_loss=0.101, over 8579.00 frames. ], tot_loss[loss=0.2435, simple_loss=0.3177, pruned_loss=0.08465, over 1163516.27 frames. ], batch size: 31, lr: 7.87e-03, grad_scale: 8.0 +2023-02-06 08:08:47,852 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-06 08:08:56,879 WARNING [train.py:1067] (0/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-06 08:09:12,574 INFO [optim.py:369] (0/4) 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 08:09:16,056 INFO [train.py:901] (0/4) Epoch 10, batch 300, loss[loss=0.3038, simple_loss=0.3592, pruned_loss=0.1242, over 7101.00 frames. ], tot_loss[loss=0.2443, simple_loss=0.3186, pruned_loss=0.085, over 1266150.07 frames. ], batch size: 72, lr: 7.87e-03, grad_scale: 8.0 +2023-02-06 08:09:23,794 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73059.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:09:40,906 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=73084.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:09:51,622 INFO [train.py:901] (0/4) Epoch 10, batch 350, loss[loss=0.2292, simple_loss=0.3194, pruned_loss=0.06953, over 8323.00 frames. ], tot_loss[loss=0.2432, simple_loss=0.3174, pruned_loss=0.08449, over 1343855.71 frames. ], batch size: 26, lr: 7.86e-03, grad_scale: 16.0 +2023-02-06 08:10:07,137 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.3686, 1.1383, 4.5366, 1.7182, 4.0109, 3.7976, 4.0634, 3.9117], + device='cuda:0'), covar=tensor([0.0427, 0.4348, 0.0388, 0.3182, 0.0992, 0.0747, 0.0471, 0.0584], + device='cuda:0'), in_proj_covar=tensor([0.0444, 0.0558, 0.0554, 0.0517, 0.0587, 0.0501, 0.0489, 0.0552], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 08:10:23,510 INFO [optim.py:369] (0/4) 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 08:10:26,925 INFO [train.py:901] (0/4) Epoch 10, batch 400, loss[loss=0.2338, simple_loss=0.3103, pruned_loss=0.07868, over 7646.00 frames. ], tot_loss[loss=0.2425, simple_loss=0.3164, pruned_loss=0.08435, over 1403413.74 frames. ], batch size: 19, lr: 7.86e-03, grad_scale: 16.0 +2023-02-06 08:11:01,346 INFO [train.py:901] (0/4) Epoch 10, batch 450, loss[loss=0.263, simple_loss=0.3342, pruned_loss=0.09588, over 8601.00 frames. ], tot_loss[loss=0.245, simple_loss=0.3182, pruned_loss=0.08584, over 1452681.27 frames. ], batch size: 31, lr: 7.86e-03, grad_scale: 16.0 +2023-02-06 08:11:33,882 INFO [optim.py:369] (0/4) 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 08:11:37,159 INFO [train.py:901] (0/4) Epoch 10, batch 500, loss[loss=0.2526, simple_loss=0.3191, pruned_loss=0.09307, over 7149.00 frames. ], tot_loss[loss=0.2452, simple_loss=0.3183, pruned_loss=0.08606, over 1488469.40 frames. ], batch size: 71, lr: 7.86e-03, grad_scale: 16.0 +2023-02-06 08:11:42,539 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=73256.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:12:10,506 INFO [train.py:901] (0/4) Epoch 10, batch 550, loss[loss=0.2661, simple_loss=0.3289, pruned_loss=0.1016, over 8368.00 frames. ], tot_loss[loss=0.2439, simple_loss=0.3167, pruned_loss=0.08559, over 1511036.56 frames. ], batch size: 24, lr: 7.85e-03, grad_scale: 16.0 +2023-02-06 08:12:19,349 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73311.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:12:21,520 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.87 vs. limit=2.0 +2023-02-06 08:12:29,143 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) Epoch 10, batch 600, loss[loss=0.2561, simple_loss=0.3241, pruned_loss=0.094, over 8510.00 frames. ], tot_loss[loss=0.244, simple_loss=0.3166, pruned_loss=0.08569, over 1534006.27 frames. ], batch size: 26, lr: 7.85e-03, grad_scale: 16.0 +2023-02-06 08:12:56,213 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-06 08:13:01,733 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=73371.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:13:20,018 INFO [train.py:901] (0/4) Epoch 10, batch 650, loss[loss=0.2072, simple_loss=0.2815, pruned_loss=0.06647, over 7672.00 frames. ], tot_loss[loss=0.2445, simple_loss=0.3169, pruned_loss=0.08603, over 1550792.13 frames. ], batch size: 18, lr: 7.85e-03, grad_scale: 16.0 +2023-02-06 08:13:50,101 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=73441.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 08:13:51,168 INFO [optim.py:369] (0/4) 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 08:13:54,595 INFO [train.py:901] (0/4) Epoch 10, batch 700, loss[loss=0.3072, simple_loss=0.361, pruned_loss=0.1267, over 6957.00 frames. ], tot_loss[loss=0.2444, simple_loss=0.317, pruned_loss=0.0859, over 1558883.23 frames. ], batch size: 71, lr: 7.84e-03, grad_scale: 16.0 +2023-02-06 08:13:55,439 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4674, 1.4346, 4.4460, 1.6711, 2.4417, 5.2171, 4.9936, 4.4690], + device='cuda:0'), covar=tensor([0.1062, 0.1787, 0.0271, 0.2004, 0.1009, 0.0158, 0.0348, 0.0577], + device='cuda:0'), in_proj_covar=tensor([0.0254, 0.0289, 0.0253, 0.0282, 0.0264, 0.0230, 0.0313, 0.0290], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 08:14:31,488 INFO [train.py:901] (0/4) Epoch 10, batch 750, loss[loss=0.2349, simple_loss=0.3147, pruned_loss=0.0775, over 8578.00 frames. ], tot_loss[loss=0.2436, simple_loss=0.3164, pruned_loss=0.08542, over 1570722.66 frames. ], batch size: 49, lr: 7.84e-03, grad_scale: 16.0 +2023-02-06 08:14:33,783 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-02-06 08:14:45,806 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-06 08:14:54,768 WARNING [train.py:1067] (0/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-06 08:15:02,255 INFO [optim.py:369] (0/4) 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 08:15:02,474 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.1591, 1.9785, 2.1577, 2.0899, 1.2063, 2.0775, 2.3214, 2.4121], + device='cuda:0'), covar=tensor([0.0421, 0.1101, 0.1577, 0.1182, 0.0576, 0.1258, 0.0608, 0.0491], + device='cuda:0'), in_proj_covar=tensor([0.0105, 0.0158, 0.0198, 0.0162, 0.0108, 0.0167, 0.0120, 0.0141], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0006], + device='cuda:0') +2023-02-06 08:15:05,724 INFO [train.py:901] (0/4) Epoch 10, batch 800, loss[loss=0.2589, simple_loss=0.3315, pruned_loss=0.09318, over 8099.00 frames. ], tot_loss[loss=0.2434, simple_loss=0.3159, pruned_loss=0.08542, over 1580647.24 frames. ], batch size: 23, lr: 7.84e-03, grad_scale: 16.0 +2023-02-06 08:15:41,996 INFO [train.py:901] (0/4) Epoch 10, batch 850, loss[loss=0.2244, simple_loss=0.308, pruned_loss=0.07043, over 8190.00 frames. ], tot_loss[loss=0.2428, simple_loss=0.3155, pruned_loss=0.08504, over 1587988.47 frames. ], batch size: 23, lr: 7.84e-03, grad_scale: 16.0 +2023-02-06 08:15:45,293 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 +2023-02-06 08:15:47,435 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 +2023-02-06 08:15:49,897 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73608.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:16:02,993 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73627.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:16:13,772 INFO [optim.py:369] (0/4) 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 08:16:17,271 INFO [train.py:901] (0/4) Epoch 10, batch 900, loss[loss=0.2301, simple_loss=0.291, pruned_loss=0.08456, over 7425.00 frames. ], tot_loss[loss=0.243, simple_loss=0.316, pruned_loss=0.08505, over 1597020.60 frames. ], batch size: 17, lr: 7.83e-03, grad_scale: 16.0 +2023-02-06 08:16:20,245 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=73652.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:16:22,229 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=73655.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:16:31,260 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1459, 1.5826, 3.3738, 1.3153, 2.1840, 3.7129, 3.6624, 3.1245], + device='cuda:0'), covar=tensor([0.0952, 0.1394, 0.0335, 0.2144, 0.0968, 0.0217, 0.0480, 0.0617], + device='cuda:0'), in_proj_covar=tensor([0.0257, 0.0291, 0.0254, 0.0282, 0.0267, 0.0232, 0.0315, 0.0289], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 08:16:52,147 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73697.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 08:16:52,449 INFO [train.py:901] (0/4) Epoch 10, batch 950, loss[loss=0.4566, simple_loss=0.4671, pruned_loss=0.2231, over 7064.00 frames. ], tot_loss[loss=0.2456, simple_loss=0.3182, pruned_loss=0.08646, over 1607638.75 frames. ], batch size: 72, lr: 7.83e-03, grad_scale: 8.0 +2023-02-06 08:17:10,363 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=73722.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 08:17:12,004 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 +2023-02-06 08:17:18,320 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-06 08:17:20,528 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2433, 2.7129, 3.1504, 1.1664, 3.0815, 2.0049, 1.5430, 1.8743], + device='cuda:0'), covar=tensor([0.0536, 0.0260, 0.0186, 0.0533, 0.0337, 0.0542, 0.0561, 0.0377], + device='cuda:0'), in_proj_covar=tensor([0.0371, 0.0305, 0.0256, 0.0360, 0.0288, 0.0454, 0.0340, 0.0332], + device='cuda:0'), out_proj_covar=tensor([1.0883e-04, 8.6973e-05, 7.3696e-05, 1.0355e-04, 8.3810e-05, 1.4272e-04, + 1.0006e-04, 9.7171e-05], device='cuda:0') +2023-02-06 08:17:24,917 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 10, batch 1000, loss[loss=0.2276, simple_loss=0.2974, pruned_loss=0.07886, over 8300.00 frames. ], tot_loss[loss=0.2448, simple_loss=0.3178, pruned_loss=0.0859, over 1613600.64 frames. ], batch size: 23, lr: 7.83e-03, grad_scale: 8.0 +2023-02-06 08:17:28,913 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73750.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:17:42,250 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=73770.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:17:50,787 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-06 08:18:00,668 INFO [train.py:901] (0/4) Epoch 10, batch 1050, loss[loss=0.2637, simple_loss=0.3363, pruned_loss=0.09551, over 8238.00 frames. ], tot_loss[loss=0.246, simple_loss=0.3188, pruned_loss=0.08662, over 1615294.77 frames. ], batch size: 24, lr: 7.83e-03, grad_scale: 8.0 +2023-02-06 08:18:01,394 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-06 08:18:34,181 INFO [optim.py:369] (0/4) 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 08:18:36,803 INFO [train.py:901] (0/4) Epoch 10, batch 1100, loss[loss=0.2701, simple_loss=0.3412, pruned_loss=0.09952, over 8135.00 frames. ], tot_loss[loss=0.2446, simple_loss=0.3178, pruned_loss=0.08569, over 1612802.32 frames. ], batch size: 22, lr: 7.82e-03, grad_scale: 8.0 +2023-02-06 08:19:09,829 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-06 08:19:10,509 INFO [train.py:901] (0/4) Epoch 10, batch 1150, loss[loss=0.2717, simple_loss=0.3554, pruned_loss=0.09401, over 8188.00 frames. ], tot_loss[loss=0.2445, simple_loss=0.318, pruned_loss=0.08543, over 1617568.26 frames. ], batch size: 23, lr: 7.82e-03, grad_scale: 8.0 +2023-02-06 08:19:35,224 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9214, 1.5080, 1.6638, 1.2118, 0.9797, 1.3588, 1.5806, 1.4408], + device='cuda:0'), covar=tensor([0.0483, 0.1201, 0.1620, 0.1369, 0.0546, 0.1481, 0.0637, 0.0585], + device='cuda:0'), in_proj_covar=tensor([0.0104, 0.0157, 0.0195, 0.0161, 0.0107, 0.0167, 0.0119, 0.0139], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0006, 0.0006], + device='cuda:0') +2023-02-06 08:19:42,441 INFO [optim.py:369] (0/4) 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 08:19:43,299 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7351, 1.9078, 2.3717, 1.6558, 1.1874, 2.5587, 0.4732, 1.3137], + device='cuda:0'), covar=tensor([0.2572, 0.1852, 0.0500, 0.2247, 0.5107, 0.0415, 0.3892, 0.2316], + device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0162, 0.0095, 0.0213, 0.0252, 0.0097, 0.0159, 0.0161], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 08:19:45,127 INFO [train.py:901] (0/4) Epoch 10, batch 1200, loss[loss=0.2455, simple_loss=0.3084, pruned_loss=0.09128, over 8091.00 frames. ], tot_loss[loss=0.2432, simple_loss=0.317, pruned_loss=0.08465, over 1614826.01 frames. ], batch size: 21, lr: 7.82e-03, grad_scale: 8.0 +2023-02-06 08:19:48,551 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=73952.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:20:20,054 INFO [train.py:901] (0/4) Epoch 10, batch 1250, loss[loss=0.2542, simple_loss=0.3369, pruned_loss=0.08572, over 8256.00 frames. ], tot_loss[loss=0.2429, simple_loss=0.3167, pruned_loss=0.0845, over 1615796.04 frames. ], batch size: 24, lr: 7.82e-03, grad_scale: 8.0 +2023-02-06 08:20:21,454 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-74000.pt +2023-02-06 08:20:39,875 INFO [zipformer.py:1185] (0/4) 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] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.58 vs. limit=5.0 +2023-02-06 08:20:51,572 INFO [optim.py:369] (0/4) 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 08:20:54,943 INFO [train.py:901] (0/4) Epoch 10, batch 1300, loss[loss=0.2125, simple_loss=0.2831, pruned_loss=0.07097, over 7800.00 frames. ], tot_loss[loss=0.2433, simple_loss=0.3167, pruned_loss=0.08495, over 1618439.24 frames. ], batch size: 19, lr: 7.81e-03, grad_scale: 8.0 +2023-02-06 08:20:57,198 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74051.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:20:58,730 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-02-06 08:21:07,838 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74067.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:21:19,112 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74082.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:21:21,532 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.92 vs. limit=5.0 +2023-02-06 08:21:26,833 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 10, batch 1350, loss[loss=0.2568, simple_loss=0.3206, pruned_loss=0.09653, over 8091.00 frames. ], tot_loss[loss=0.2473, simple_loss=0.3204, pruned_loss=0.08706, over 1625716.79 frames. ], batch size: 21, lr: 7.81e-03, grad_scale: 8.0 +2023-02-06 08:21:59,543 INFO [optim.py:369] (0/4) 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 08:22:02,259 INFO [train.py:901] (0/4) Epoch 10, batch 1400, loss[loss=0.2562, simple_loss=0.3204, pruned_loss=0.09601, over 7522.00 frames. ], tot_loss[loss=0.2484, simple_loss=0.3211, pruned_loss=0.08791, over 1624933.15 frames. ], batch size: 18, lr: 7.81e-03, grad_scale: 8.0 +2023-02-06 08:22:16,762 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 +2023-02-06 08:22:38,035 INFO [train.py:901] (0/4) Epoch 10, batch 1450, loss[loss=0.1913, simple_loss=0.2688, pruned_loss=0.05695, over 7802.00 frames. ], tot_loss[loss=0.2457, simple_loss=0.3187, pruned_loss=0.0864, over 1623161.63 frames. ], batch size: 19, lr: 7.81e-03, grad_scale: 8.0 +2023-02-06 08:22:41,671 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-06 08:22:45,869 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74209.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:23:01,273 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.6033, 4.6147, 4.1190, 1.8530, 4.0756, 4.2837, 4.1993, 3.8249], + device='cuda:0'), covar=tensor([0.0572, 0.0484, 0.0892, 0.4715, 0.0756, 0.0688, 0.1192, 0.0751], + device='cuda:0'), in_proj_covar=tensor([0.0448, 0.0356, 0.0377, 0.0470, 0.0372, 0.0353, 0.0361, 0.0314], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 08:23:01,603 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 +2023-02-06 08:23:07,376 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9898, 1.5568, 3.0047, 1.2141, 2.1456, 3.3188, 3.3752, 2.7998], + device='cuda:0'), covar=tensor([0.0982, 0.1489, 0.0424, 0.2125, 0.0970, 0.0271, 0.0460, 0.0682], + device='cuda:0'), in_proj_covar=tensor([0.0256, 0.0293, 0.0254, 0.0284, 0.0269, 0.0232, 0.0318, 0.0289], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 08:23:09,187 INFO [optim.py:369] (0/4) 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 08:23:11,846 INFO [train.py:901] (0/4) Epoch 10, batch 1500, loss[loss=0.189, simple_loss=0.2643, pruned_loss=0.0569, over 7412.00 frames. ], tot_loss[loss=0.244, simple_loss=0.3175, pruned_loss=0.08529, over 1620239.00 frames. ], batch size: 17, lr: 7.80e-03, grad_scale: 8.0 +2023-02-06 08:23:18,346 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74258.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:23:46,603 INFO [train.py:901] (0/4) Epoch 10, batch 1550, loss[loss=0.2255, simple_loss=0.3046, pruned_loss=0.07315, over 8101.00 frames. ], tot_loss[loss=0.2437, simple_loss=0.3173, pruned_loss=0.08507, over 1618445.84 frames. ], batch size: 23, lr: 7.80e-03, grad_scale: 8.0 +2023-02-06 08:23:47,771 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-02-06 08:24:05,673 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74323.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:24:19,877 INFO [optim.py:369] (0/4) 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 08:24:22,705 INFO [train.py:901] (0/4) Epoch 10, batch 1600, loss[loss=0.2142, simple_loss=0.2865, pruned_loss=0.07093, over 7787.00 frames. ], tot_loss[loss=0.2437, simple_loss=0.3173, pruned_loss=0.08506, over 1616956.92 frames. ], batch size: 19, lr: 7.80e-03, grad_scale: 8.0 +2023-02-06 08:24:22,901 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74348.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:24:56,947 INFO [train.py:901] (0/4) Epoch 10, batch 1650, loss[loss=0.239, simple_loss=0.3043, pruned_loss=0.08684, over 7526.00 frames. ], tot_loss[loss=0.2437, simple_loss=0.3171, pruned_loss=0.08514, over 1617053.47 frames. ], batch size: 18, lr: 7.79e-03, grad_scale: 8.0 +2023-02-06 08:25:18,074 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) attn_weights_entropy = tensor([1.2994, 1.4725, 1.2343, 1.8688, 0.7497, 1.0048, 1.2470, 1.4034], + device='cuda:0'), covar=tensor([0.1028, 0.0935, 0.1301, 0.0603, 0.1330, 0.1925, 0.0991, 0.0912], + device='cuda:0'), in_proj_covar=tensor([0.0241, 0.0216, 0.0261, 0.0217, 0.0221, 0.0255, 0.0261, 0.0222], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 08:25:32,856 INFO [train.py:901] (0/4) Epoch 10, batch 1700, loss[loss=0.2491, simple_loss=0.3267, pruned_loss=0.08576, over 8460.00 frames. ], tot_loss[loss=0.2427, simple_loss=0.3169, pruned_loss=0.08423, over 1620782.86 frames. ], batch size: 25, lr: 7.79e-03, grad_scale: 8.0 +2023-02-06 08:25:44,193 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74490.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:26:05,897 INFO [train.py:901] (0/4) Epoch 10, batch 1750, loss[loss=0.1857, simple_loss=0.2712, pruned_loss=0.05004, over 8241.00 frames. ], tot_loss[loss=0.2441, simple_loss=0.3178, pruned_loss=0.08525, over 1620091.83 frames. ], batch size: 22, lr: 7.79e-03, grad_scale: 8.0 +2023-02-06 08:26:36,848 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74541.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:26:38,772 INFO [optim.py:369] (0/4) 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 08:26:41,525 INFO [train.py:901] (0/4) Epoch 10, batch 1800, loss[loss=0.2436, simple_loss=0.3209, pruned_loss=0.08319, over 8249.00 frames. ], tot_loss[loss=0.2451, simple_loss=0.3185, pruned_loss=0.08583, over 1622644.78 frames. ], batch size: 24, lr: 7.79e-03, grad_scale: 8.0 +2023-02-06 08:27:14,907 INFO [train.py:901] (0/4) Epoch 10, batch 1850, loss[loss=0.1986, simple_loss=0.2714, pruned_loss=0.06288, over 7436.00 frames. ], tot_loss[loss=0.2451, simple_loss=0.3181, pruned_loss=0.08602, over 1621981.83 frames. ], batch size: 17, lr: 7.78e-03, grad_scale: 8.0 +2023-02-06 08:27:17,782 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=74602.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:27:25,264 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74613.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:27:46,970 INFO [optim.py:369] (0/4) 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 08:27:50,399 INFO [train.py:901] (0/4) Epoch 10, batch 1900, loss[loss=0.2499, simple_loss=0.316, pruned_loss=0.0919, over 8080.00 frames. ], tot_loss[loss=0.2449, simple_loss=0.3183, pruned_loss=0.0857, over 1620711.93 frames. ], batch size: 21, lr: 7.78e-03, grad_scale: 8.0 +2023-02-06 08:28:13,885 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-06 08:28:25,273 INFO [train.py:901] (0/4) Epoch 10, batch 1950, loss[loss=0.1955, simple_loss=0.2689, pruned_loss=0.06102, over 7419.00 frames. ], tot_loss[loss=0.2434, simple_loss=0.3172, pruned_loss=0.0848, over 1616851.44 frames. ], batch size: 17, lr: 7.78e-03, grad_scale: 8.0 +2023-02-06 08:28:25,967 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-06 08:28:38,213 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74717.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:28:41,921 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 +2023-02-06 08:28:43,969 WARNING [train.py:1067] (0/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-06 08:28:56,745 INFO [optim.py:369] (0/4) 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 08:28:59,513 INFO [train.py:901] (0/4) Epoch 10, batch 2000, loss[loss=0.238, simple_loss=0.3108, pruned_loss=0.08267, over 8255.00 frames. ], tot_loss[loss=0.244, simple_loss=0.3181, pruned_loss=0.08495, over 1619357.99 frames. ], batch size: 22, lr: 7.78e-03, grad_scale: 8.0 +2023-02-06 08:29:34,146 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74797.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:29:34,597 INFO [train.py:901] (0/4) Epoch 10, batch 2050, loss[loss=0.2286, simple_loss=0.3067, pruned_loss=0.07529, over 8248.00 frames. ], tot_loss[loss=0.2424, simple_loss=0.3163, pruned_loss=0.08426, over 1619462.22 frames. ], batch size: 24, lr: 7.77e-03, grad_scale: 8.0 +2023-02-06 08:29:50,346 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74822.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:29:58,215 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3649, 1.2060, 1.5526, 1.2173, 0.7140, 1.3437, 1.2640, 1.0860], + device='cuda:0'), covar=tensor([0.0544, 0.1347, 0.1747, 0.1490, 0.0588, 0.1560, 0.0686, 0.0651], + device='cuda:0'), in_proj_covar=tensor([0.0103, 0.0156, 0.0197, 0.0162, 0.0106, 0.0167, 0.0119, 0.0139], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0006, 0.0006], + device='cuda:0') +2023-02-06 08:30:04,620 INFO [optim.py:369] (0/4) 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 08:30:06,169 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([0.9993, 1.6365, 1.3361, 1.5523, 1.3761, 1.1021, 1.1472, 1.3417], + device='cuda:0'), covar=tensor([0.0949, 0.0377, 0.1076, 0.0467, 0.0643, 0.1326, 0.0826, 0.0683], + device='cuda:0'), in_proj_covar=tensor([0.0342, 0.0232, 0.0312, 0.0296, 0.0304, 0.0317, 0.0337, 0.0308], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 08:30:07,354 INFO [train.py:901] (0/4) Epoch 10, batch 2100, loss[loss=0.2267, simple_loss=0.309, pruned_loss=0.0722, over 8249.00 frames. ], tot_loss[loss=0.2441, simple_loss=0.3179, pruned_loss=0.08514, over 1617856.43 frames. ], batch size: 24, lr: 7.77e-03, grad_scale: 8.0 +2023-02-06 08:30:18,141 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6349, 1.5917, 1.9545, 1.5310, 0.9719, 2.0252, 0.1850, 1.1541], + device='cuda:0'), covar=tensor([0.2821, 0.2089, 0.0498, 0.2148, 0.4950, 0.0461, 0.3993, 0.2136], + device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0163, 0.0092, 0.0213, 0.0254, 0.0097, 0.0161, 0.0158], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0001, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 08:30:21,942 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.5722, 4.6294, 4.1065, 1.9835, 4.1288, 4.1166, 4.1924, 3.8483], + device='cuda:0'), covar=tensor([0.0801, 0.0539, 0.1081, 0.4912, 0.0853, 0.0870, 0.1134, 0.0802], + device='cuda:0'), in_proj_covar=tensor([0.0444, 0.0358, 0.0379, 0.0465, 0.0367, 0.0352, 0.0361, 0.0315], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 08:30:24,010 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4444, 1.8296, 1.7953, 1.0579, 1.9432, 1.4277, 0.4250, 1.6220], + device='cuda:0'), covar=tensor([0.0307, 0.0171, 0.0197, 0.0315, 0.0238, 0.0515, 0.0495, 0.0142], + device='cuda:0'), in_proj_covar=tensor([0.0374, 0.0306, 0.0265, 0.0364, 0.0296, 0.0452, 0.0348, 0.0332], + device='cuda:0'), out_proj_covar=tensor([1.0950e-04, 8.7103e-05, 7.5990e-05, 1.0479e-04, 8.5882e-05, 1.4168e-04, + 1.0204e-04, 9.6580e-05], device='cuda:0') +2023-02-06 08:30:43,220 INFO [train.py:901] (0/4) Epoch 10, batch 2150, loss[loss=0.2625, simple_loss=0.3331, pruned_loss=0.0959, over 8196.00 frames. ], tot_loss[loss=0.2433, simple_loss=0.3169, pruned_loss=0.08483, over 1615901.20 frames. ], batch size: 23, lr: 7.77e-03, grad_scale: 8.0 +2023-02-06 08:31:02,633 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 08:31:16,706 INFO [train.py:901] (0/4) Epoch 10, batch 2200, loss[loss=0.2525, simple_loss=0.3264, pruned_loss=0.08927, over 8349.00 frames. ], tot_loss[loss=0.2421, simple_loss=0.3161, pruned_loss=0.08411, over 1613079.58 frames. ], batch size: 24, lr: 7.77e-03, grad_scale: 8.0 +2023-02-06 08:31:22,705 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=74957.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:31:33,576 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74973.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:31:39,398 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74982.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:31:50,363 INFO [train.py:901] (0/4) Epoch 10, batch 2250, loss[loss=0.2276, simple_loss=0.3108, pruned_loss=0.07222, over 8135.00 frames. ], tot_loss[loss=0.2422, simple_loss=0.316, pruned_loss=0.0842, over 1615525.85 frames. ], batch size: 22, lr: 7.76e-03, grad_scale: 8.0 +2023-02-06 08:31:50,574 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74998.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:32:21,307 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7321, 2.3687, 1.6421, 2.7001, 1.3929, 1.2706, 2.0924, 2.3122], + device='cuda:0'), covar=tensor([0.1002, 0.0780, 0.1426, 0.0469, 0.1168, 0.1851, 0.0915, 0.0763], + device='cuda:0'), in_proj_covar=tensor([0.0244, 0.0216, 0.0262, 0.0218, 0.0223, 0.0255, 0.0263, 0.0223], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 08:32:23,118 INFO [optim.py:369] (0/4) 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 08:32:25,870 INFO [train.py:901] (0/4) Epoch 10, batch 2300, loss[loss=0.1994, simple_loss=0.2824, pruned_loss=0.05815, over 8249.00 frames. ], tot_loss[loss=0.2437, simple_loss=0.317, pruned_loss=0.08514, over 1619217.36 frames. ], batch size: 24, lr: 7.76e-03, grad_scale: 8.0 +2023-02-06 08:32:42,300 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 10, batch 2350, loss[loss=0.2237, simple_loss=0.2893, pruned_loss=0.07908, over 7654.00 frames. ], tot_loss[loss=0.2431, simple_loss=0.3164, pruned_loss=0.08493, over 1615624.09 frames. ], batch size: 19, lr: 7.76e-03, grad_scale: 8.0 +2023-02-06 08:33:17,158 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.2077, 4.2201, 3.8191, 1.7787, 3.7731, 3.7549, 3.8986, 3.4729], + device='cuda:0'), covar=tensor([0.0788, 0.0568, 0.0925, 0.4339, 0.0768, 0.0968, 0.1013, 0.0836], + device='cuda:0'), in_proj_covar=tensor([0.0452, 0.0362, 0.0383, 0.0472, 0.0371, 0.0354, 0.0366, 0.0317], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 08:33:33,027 INFO [optim.py:369] (0/4) 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 08:33:35,802 INFO [train.py:901] (0/4) Epoch 10, batch 2400, loss[loss=0.2257, simple_loss=0.3086, pruned_loss=0.07143, over 8104.00 frames. ], tot_loss[loss=0.243, simple_loss=0.3163, pruned_loss=0.08486, over 1617261.49 frames. ], batch size: 23, lr: 7.76e-03, grad_scale: 8.0 +2023-02-06 08:33:40,204 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 +2023-02-06 08:34:08,760 INFO [train.py:901] (0/4) Epoch 10, batch 2450, loss[loss=0.2258, simple_loss=0.2921, pruned_loss=0.07977, over 8243.00 frames. ], tot_loss[loss=0.2435, simple_loss=0.3164, pruned_loss=0.08534, over 1617670.86 frames. ], batch size: 22, lr: 7.75e-03, grad_scale: 8.0 +2023-02-06 08:34:26,434 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 +2023-02-06 08:34:40,829 INFO [optim.py:369] (0/4) 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 08:34:43,537 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=75246.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:34:44,707 INFO [train.py:901] (0/4) Epoch 10, batch 2500, loss[loss=0.235, simple_loss=0.3066, pruned_loss=0.08168, over 8325.00 frames. ], tot_loss[loss=0.2438, simple_loss=0.3166, pruned_loss=0.08552, over 1618669.00 frames. ], batch size: 25, lr: 7.75e-03, grad_scale: 8.0 +2023-02-06 08:35:00,228 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=75271.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:35:11,669 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 10, batch 2550, loss[loss=0.1803, simple_loss=0.2543, pruned_loss=0.0531, over 7543.00 frames. ], tot_loss[loss=0.243, simple_loss=0.3165, pruned_loss=0.08476, over 1623007.91 frames. ], batch size: 18, lr: 7.75e-03, grad_scale: 8.0 +2023-02-06 08:35:36,578 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=75326.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:35:38,121 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=75328.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:35:49,134 INFO [optim.py:369] (0/4) 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 08:35:51,836 INFO [train.py:901] (0/4) Epoch 10, batch 2600, loss[loss=0.2523, simple_loss=0.3229, pruned_loss=0.09083, over 8571.00 frames. ], tot_loss[loss=0.2431, simple_loss=0.3169, pruned_loss=0.08467, over 1627938.05 frames. ], batch size: 31, lr: 7.75e-03, grad_scale: 8.0 +2023-02-06 08:35:55,406 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=75353.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:36:19,394 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75386.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:36:27,351 INFO [train.py:901] (0/4) Epoch 10, batch 2650, loss[loss=0.1974, simple_loss=0.2969, pruned_loss=0.04896, over 8472.00 frames. ], tot_loss[loss=0.2427, simple_loss=0.3162, pruned_loss=0.08463, over 1623415.07 frames. ], batch size: 25, lr: 7.74e-03, grad_scale: 8.0 +2023-02-06 08:36:52,096 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0358, 1.6530, 1.3355, 1.5725, 1.3245, 1.2056, 1.2616, 1.4238], + device='cuda:0'), covar=tensor([0.0935, 0.0385, 0.1042, 0.0477, 0.0693, 0.1251, 0.0744, 0.0684], + device='cuda:0'), in_proj_covar=tensor([0.0345, 0.0235, 0.0313, 0.0299, 0.0305, 0.0319, 0.0340, 0.0312], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 08:36:55,407 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.2668, 1.3674, 4.5849, 1.9016, 3.6290, 3.6519, 4.0914, 4.0742], + device='cuda:0'), covar=tensor([0.0961, 0.6342, 0.0922, 0.4342, 0.2237, 0.1420, 0.0991, 0.0999], + device='cuda:0'), in_proj_covar=tensor([0.0452, 0.0560, 0.0566, 0.0516, 0.0593, 0.0506, 0.0494, 0.0555], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 08:36:56,747 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75441.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:36:58,552 INFO [optim.py:369] (0/4) 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 08:37:01,350 INFO [train.py:901] (0/4) Epoch 10, batch 2700, loss[loss=0.2028, simple_loss=0.2871, pruned_loss=0.05925, over 8187.00 frames. ], tot_loss[loss=0.2424, simple_loss=0.3158, pruned_loss=0.08453, over 1622645.02 frames. ], batch size: 23, lr: 7.74e-03, grad_scale: 8.0 +2023-02-06 08:37:01,849 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 +2023-02-06 08:37:37,693 INFO [train.py:901] (0/4) Epoch 10, batch 2750, loss[loss=0.2727, simple_loss=0.3317, pruned_loss=0.1069, over 8500.00 frames. ], tot_loss[loss=0.2414, simple_loss=0.3146, pruned_loss=0.0841, over 1617162.26 frames. ], batch size: 39, lr: 7.74e-03, grad_scale: 8.0 +2023-02-06 08:38:01,209 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.84 vs. limit=5.0 +2023-02-06 08:38:08,180 INFO [optim.py:369] (0/4) 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 08:38:10,729 INFO [train.py:901] (0/4) Epoch 10, batch 2800, loss[loss=0.2891, simple_loss=0.3646, pruned_loss=0.1068, over 8025.00 frames. ], tot_loss[loss=0.2404, simple_loss=0.3136, pruned_loss=0.08363, over 1610036.63 frames. ], batch size: 22, lr: 7.74e-03, grad_scale: 8.0 +2023-02-06 08:38:15,241 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 +2023-02-06 08:38:39,095 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=75590.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:38:45,071 INFO [train.py:901] (0/4) Epoch 10, batch 2850, loss[loss=0.2244, simple_loss=0.3128, pruned_loss=0.06802, over 8485.00 frames. ], tot_loss[loss=0.2399, simple_loss=0.3132, pruned_loss=0.08329, over 1610868.66 frames. ], batch size: 29, lr: 7.73e-03, grad_scale: 8.0 +2023-02-06 08:38:52,328 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 +2023-02-06 08:39:06,714 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1995, 1.4678, 4.6335, 1.8909, 2.7654, 5.2370, 5.0747, 4.5398], + device='cuda:0'), covar=tensor([0.1111, 0.1749, 0.0222, 0.1861, 0.0850, 0.0147, 0.0307, 0.0556], + device='cuda:0'), in_proj_covar=tensor([0.0258, 0.0293, 0.0253, 0.0284, 0.0268, 0.0233, 0.0319, 0.0289], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 08:39:09,360 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=75632.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:39:09,745 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 +2023-02-06 08:39:16,271 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) Epoch 10, batch 2900, loss[loss=0.2493, simple_loss=0.3205, pruned_loss=0.08911, over 8445.00 frames. ], tot_loss[loss=0.2422, simple_loss=0.3151, pruned_loss=0.08469, over 1612182.07 frames. ], batch size: 27, lr: 7.73e-03, grad_scale: 8.0 +2023-02-06 08:39:32,775 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=75667.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:39:49,791 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-06 08:39:53,279 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=75697.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:39:53,742 INFO [train.py:901] (0/4) Epoch 10, batch 2950, loss[loss=0.2576, simple_loss=0.3313, pruned_loss=0.09195, over 8678.00 frames. ], tot_loss[loss=0.2427, simple_loss=0.3156, pruned_loss=0.08487, over 1610866.32 frames. ], batch size: 39, lr: 7.73e-03, grad_scale: 16.0 +2023-02-06 08:39:58,746 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75705.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:40:10,317 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 +2023-02-06 08:40:11,383 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=75722.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:40:26,680 INFO [optim.py:369] (0/4) 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 08:40:28,944 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75747.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:40:29,494 INFO [train.py:901] (0/4) Epoch 10, batch 3000, loss[loss=0.1964, simple_loss=0.2782, pruned_loss=0.05731, over 8089.00 frames. ], tot_loss[loss=0.2436, simple_loss=0.3165, pruned_loss=0.0853, over 1614631.82 frames. ], batch size: 21, lr: 7.73e-03, grad_scale: 16.0 +2023-02-06 08:40:29,494 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 08:40:41,876 INFO [train.py:935] (0/4) Epoch 10, validation: loss=0.1918, simple_loss=0.2916, pruned_loss=0.04599, over 944034.00 frames. +2023-02-06 08:40:41,877 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-06 08:41:15,537 INFO [train.py:901] (0/4) Epoch 10, batch 3050, loss[loss=0.227, simple_loss=0.308, pruned_loss=0.07302, over 8113.00 frames. ], tot_loss[loss=0.2454, simple_loss=0.3182, pruned_loss=0.08634, over 1619423.78 frames. ], batch size: 23, lr: 7.72e-03, grad_scale: 16.0 +2023-02-06 08:41:47,920 INFO [optim.py:369] (0/4) 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 08:41:50,031 INFO [train.py:901] (0/4) Epoch 10, batch 3100, loss[loss=0.2132, simple_loss=0.2799, pruned_loss=0.07318, over 7701.00 frames. ], tot_loss[loss=0.2455, simple_loss=0.3186, pruned_loss=0.08625, over 1614848.46 frames. ], batch size: 18, lr: 7.72e-03, grad_scale: 8.0 +2023-02-06 08:42:19,305 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 +2023-02-06 08:42:25,560 INFO [train.py:901] (0/4) Epoch 10, batch 3150, loss[loss=0.2194, simple_loss=0.2898, pruned_loss=0.07457, over 7652.00 frames. ], tot_loss[loss=0.2457, simple_loss=0.3184, pruned_loss=0.08652, over 1609482.04 frames. ], batch size: 19, lr: 7.72e-03, grad_scale: 8.0 +2023-02-06 08:42:57,479 INFO [optim.py:369] (0/4) 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 08:42:59,542 INFO [train.py:901] (0/4) Epoch 10, batch 3200, loss[loss=0.2554, simple_loss=0.3409, pruned_loss=0.08496, over 8515.00 frames. ], tot_loss[loss=0.244, simple_loss=0.317, pruned_loss=0.08551, over 1607415.25 frames. ], batch size: 31, lr: 7.72e-03, grad_scale: 8.0 +2023-02-06 08:43:09,332 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=75961.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:43:28,328 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=75986.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:43:36,373 INFO [train.py:901] (0/4) Epoch 10, batch 3250, loss[loss=0.3027, simple_loss=0.358, pruned_loss=0.1237, over 6761.00 frames. ], tot_loss[loss=0.2436, simple_loss=0.3164, pruned_loss=0.08534, over 1603131.85 frames. ], batch size: 71, lr: 7.71e-03, grad_scale: 8.0 +2023-02-06 08:43:37,556 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-02-06 08:43:37,964 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-76000.pt +2023-02-06 08:43:41,248 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=76003.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:43:56,207 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.61 vs. limit=5.0 +2023-02-06 08:43:57,873 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=76028.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:44:09,017 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 10, batch 3300, loss[loss=0.2918, simple_loss=0.3378, pruned_loss=0.1229, over 7656.00 frames. ], tot_loss[loss=0.2451, simple_loss=0.3178, pruned_loss=0.08624, over 1608549.95 frames. ], batch size: 19, lr: 7.71e-03, grad_scale: 8.0 +2023-02-06 08:44:47,579 INFO [train.py:901] (0/4) Epoch 10, batch 3350, loss[loss=0.1931, simple_loss=0.2666, pruned_loss=0.05979, over 7801.00 frames. ], tot_loss[loss=0.2429, simple_loss=0.3159, pruned_loss=0.08493, over 1605263.83 frames. ], batch size: 20, lr: 7.71e-03, grad_scale: 8.0 +2023-02-06 08:45:18,108 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7642, 1.4977, 2.7807, 1.2450, 2.0644, 2.9701, 3.0526, 2.5260], + device='cuda:0'), covar=tensor([0.1037, 0.1346, 0.0436, 0.2045, 0.0846, 0.0316, 0.0580, 0.0759], + device='cuda:0'), in_proj_covar=tensor([0.0256, 0.0293, 0.0253, 0.0284, 0.0268, 0.0234, 0.0319, 0.0289], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 08:45:18,633 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 10, batch 3400, loss[loss=0.2342, simple_loss=0.3115, pruned_loss=0.0785, over 8510.00 frames. ], tot_loss[loss=0.2418, simple_loss=0.3153, pruned_loss=0.08416, over 1610189.01 frames. ], batch size: 28, lr: 7.71e-03, grad_scale: 8.0 +2023-02-06 08:45:55,814 INFO [train.py:901] (0/4) Epoch 10, batch 3450, loss[loss=0.2336, simple_loss=0.3206, pruned_loss=0.07326, over 7975.00 frames. ], tot_loss[loss=0.2424, simple_loss=0.316, pruned_loss=0.08444, over 1612827.34 frames. ], batch size: 21, lr: 7.70e-03, grad_scale: 8.0 +2023-02-06 08:46:30,137 INFO [optim.py:369] (0/4) 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 08:46:32,217 INFO [train.py:901] (0/4) Epoch 10, batch 3500, loss[loss=0.3005, simple_loss=0.3517, pruned_loss=0.1246, over 8514.00 frames. ], tot_loss[loss=0.2436, simple_loss=0.3167, pruned_loss=0.08526, over 1610108.39 frames. ], batch size: 49, lr: 7.70e-03, grad_scale: 8.0 +2023-02-06 08:46:39,255 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3184, 2.7399, 1.8458, 2.1991, 2.1405, 1.4637, 1.8573, 2.2475], + device='cuda:0'), covar=tensor([0.1367, 0.0326, 0.1074, 0.0580, 0.0659, 0.1415, 0.0990, 0.0790], + device='cuda:0'), in_proj_covar=tensor([0.0346, 0.0234, 0.0314, 0.0299, 0.0308, 0.0325, 0.0340, 0.0314], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 08:46:48,050 WARNING [train.py:1067] (0/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-06 08:46:51,371 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 +2023-02-06 08:46:59,271 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=76287.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 08:47:06,736 INFO [train.py:901] (0/4) Epoch 10, batch 3550, loss[loss=0.2074, simple_loss=0.2674, pruned_loss=0.07372, over 7197.00 frames. ], tot_loss[loss=0.2418, simple_loss=0.3152, pruned_loss=0.08424, over 1607160.95 frames. ], batch size: 16, lr: 7.70e-03, grad_scale: 8.0 +2023-02-06 08:47:17,280 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=76312.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:47:39,387 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=76341.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 08:47:42,026 INFO [optim.py:369] (0/4) 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 08:47:44,169 INFO [train.py:901] (0/4) Epoch 10, batch 3600, loss[loss=0.2271, simple_loss=0.2977, pruned_loss=0.07827, over 8073.00 frames. ], tot_loss[loss=0.2407, simple_loss=0.314, pruned_loss=0.08368, over 1604911.80 frames. ], batch size: 21, lr: 7.70e-03, grad_scale: 8.0 +2023-02-06 08:47:53,341 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.0259, 1.4327, 4.2893, 1.5867, 3.7221, 3.5056, 3.8377, 3.6990], + device='cuda:0'), covar=tensor([0.0640, 0.4414, 0.0500, 0.3433, 0.1249, 0.0901, 0.0590, 0.0698], + device='cuda:0'), in_proj_covar=tensor([0.0453, 0.0568, 0.0573, 0.0524, 0.0603, 0.0512, 0.0499, 0.0564], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 08:48:18,377 INFO [train.py:901] (0/4) Epoch 10, batch 3650, loss[loss=0.2543, simple_loss=0.3281, pruned_loss=0.09022, over 8502.00 frames. ], tot_loss[loss=0.24, simple_loss=0.3132, pruned_loss=0.08342, over 1602057.31 frames. ], batch size: 39, lr: 7.69e-03, grad_scale: 8.0 +2023-02-06 08:48:50,984 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 10, batch 3700, loss[loss=0.246, simple_loss=0.3272, pruned_loss=0.08241, over 8143.00 frames. ], tot_loss[loss=0.2409, simple_loss=0.3145, pruned_loss=0.08362, over 1609308.79 frames. ], batch size: 22, lr: 7.69e-03, grad_scale: 8.0 +2023-02-06 08:48:54,935 WARNING [train.py:1067] (0/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-06 08:49:28,897 INFO [train.py:901] (0/4) Epoch 10, batch 3750, loss[loss=0.2882, simple_loss=0.3475, pruned_loss=0.1145, over 6933.00 frames. ], tot_loss[loss=0.2427, simple_loss=0.3162, pruned_loss=0.08457, over 1610426.70 frames. ], batch size: 71, lr: 7.69e-03, grad_scale: 8.0 +2023-02-06 08:50:00,300 INFO [optim.py:369] (0/4) 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 08:50:02,995 INFO [train.py:901] (0/4) Epoch 10, batch 3800, loss[loss=0.2178, simple_loss=0.2946, pruned_loss=0.07051, over 7529.00 frames. ], tot_loss[loss=0.2416, simple_loss=0.3154, pruned_loss=0.0839, over 1610611.34 frames. ], batch size: 18, lr: 7.69e-03, grad_scale: 8.0 +2023-02-06 08:50:38,435 INFO [train.py:901] (0/4) Epoch 10, batch 3850, loss[loss=0.2287, simple_loss=0.3118, pruned_loss=0.07279, over 8448.00 frames. ], tot_loss[loss=0.2414, simple_loss=0.3149, pruned_loss=0.08398, over 1609506.45 frames. ], batch size: 27, lr: 7.68e-03, grad_scale: 8.0 +2023-02-06 08:50:59,075 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-06 08:51:00,498 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=76631.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 08:51:09,921 INFO [optim.py:369] (0/4) 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 08:51:11,980 INFO [train.py:901] (0/4) Epoch 10, batch 3900, loss[loss=0.2301, simple_loss=0.3178, pruned_loss=0.0712, over 8497.00 frames. ], tot_loss[loss=0.2417, simple_loss=0.3151, pruned_loss=0.08413, over 1611193.79 frames. ], batch size: 28, lr: 7.68e-03, grad_scale: 8.0 +2023-02-06 08:51:14,241 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5093, 1.7698, 2.8547, 1.3319, 1.9757, 1.8937, 1.5801, 1.7946], + device='cuda:0'), covar=tensor([0.1778, 0.2228, 0.0707, 0.3726, 0.1560, 0.2712, 0.1719, 0.2081], + device='cuda:0'), in_proj_covar=tensor([0.0482, 0.0502, 0.0529, 0.0565, 0.0605, 0.0544, 0.0463, 0.0601], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 08:51:17,431 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=76656.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:51:38,048 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=76685.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 08:51:48,068 INFO [train.py:901] (0/4) Epoch 10, batch 3950, loss[loss=0.2481, simple_loss=0.3258, pruned_loss=0.08522, over 7805.00 frames. ], tot_loss[loss=0.242, simple_loss=0.3154, pruned_loss=0.08427, over 1612281.96 frames. ], batch size: 20, lr: 7.68e-03, grad_scale: 8.0 +2023-02-06 08:51:50,990 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6292, 1.3262, 1.5324, 1.1545, 0.8764, 1.2628, 1.4266, 1.2364], + device='cuda:0'), covar=tensor([0.0564, 0.1329, 0.1869, 0.1551, 0.0635, 0.1648, 0.0759, 0.0702], + device='cuda:0'), in_proj_covar=tensor([0.0103, 0.0157, 0.0197, 0.0162, 0.0106, 0.0166, 0.0119, 0.0139], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0006, 0.0006], + device='cuda:0') +2023-02-06 08:52:19,565 INFO [optim.py:369] (0/4) 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 08:52:20,444 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 10, batch 4000, loss[loss=0.2019, simple_loss=0.2734, pruned_loss=0.06523, over 7536.00 frames. ], tot_loss[loss=0.2403, simple_loss=0.3138, pruned_loss=0.0834, over 1608746.61 frames. ], batch size: 18, lr: 7.68e-03, grad_scale: 8.0 +2023-02-06 08:52:32,357 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 +2023-02-06 08:52:36,775 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4580, 2.0158, 2.1120, 1.0555, 2.1802, 1.4201, 0.5337, 1.6678], + device='cuda:0'), covar=tensor([0.0423, 0.0204, 0.0152, 0.0372, 0.0242, 0.0644, 0.0562, 0.0189], + device='cuda:0'), in_proj_covar=tensor([0.0375, 0.0307, 0.0263, 0.0369, 0.0299, 0.0454, 0.0351, 0.0338], + device='cuda:0'), out_proj_covar=tensor([1.0894e-04, 8.7150e-05, 7.4858e-05, 1.0548e-04, 8.6795e-05, 1.4130e-04, + 1.0227e-04, 9.8215e-05], device='cuda:0') +2023-02-06 08:52:37,413 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=76771.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:52:56,078 INFO [train.py:901] (0/4) Epoch 10, batch 4050, loss[loss=0.2335, simple_loss=0.3136, pruned_loss=0.07672, over 8641.00 frames. ], tot_loss[loss=0.2401, simple_loss=0.314, pruned_loss=0.08307, over 1611832.62 frames. ], batch size: 34, lr: 7.67e-03, grad_scale: 8.0 +2023-02-06 08:52:57,688 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=76800.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 08:53:29,318 INFO [optim.py:369] (0/4) 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 08:53:31,234 INFO [train.py:901] (0/4) Epoch 10, batch 4100, loss[loss=0.2688, simple_loss=0.3356, pruned_loss=0.101, over 8016.00 frames. ], tot_loss[loss=0.2401, simple_loss=0.314, pruned_loss=0.08304, over 1611391.57 frames. ], batch size: 22, lr: 7.67e-03, grad_scale: 8.0 +2023-02-06 08:53:37,249 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 10, batch 4150, loss[loss=0.251, simple_loss=0.3405, pruned_loss=0.08071, over 8487.00 frames. ], tot_loss[loss=0.2397, simple_loss=0.3135, pruned_loss=0.08294, over 1610909.83 frames. ], batch size: 28, lr: 7.67e-03, grad_scale: 8.0 +2023-02-06 08:54:38,768 INFO [optim.py:369] (0/4) 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 08:54:40,851 INFO [train.py:901] (0/4) Epoch 10, batch 4200, loss[loss=0.2697, simple_loss=0.3487, pruned_loss=0.09538, over 8468.00 frames. ], tot_loss[loss=0.2377, simple_loss=0.3121, pruned_loss=0.08166, over 1611767.45 frames. ], batch size: 25, lr: 7.67e-03, grad_scale: 8.0 +2023-02-06 08:54:54,286 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.5653, 3.5128, 3.2130, 2.3697, 3.1601, 3.1043, 3.3644, 2.8745], + device='cuda:0'), covar=tensor([0.0862, 0.0755, 0.0952, 0.3105, 0.0864, 0.1050, 0.1123, 0.1002], + device='cuda:0'), in_proj_covar=tensor([0.0449, 0.0359, 0.0370, 0.0468, 0.0364, 0.0355, 0.0361, 0.0310], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 08:55:00,931 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-06 08:55:14,228 INFO [train.py:901] (0/4) Epoch 10, batch 4250, loss[loss=0.2363, simple_loss=0.313, pruned_loss=0.07983, over 8072.00 frames. ], tot_loss[loss=0.2379, simple_loss=0.3125, pruned_loss=0.08165, over 1614296.46 frames. ], batch size: 21, lr: 7.66e-03, grad_scale: 8.0 +2023-02-06 08:55:17,198 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77002.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 08:55:23,796 WARNING [train.py:1067] (0/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-06 08:55:34,046 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77027.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 08:55:34,056 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77027.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:55:46,498 INFO [optim.py:369] (0/4) 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 08:55:48,445 INFO [train.py:901] (0/4) Epoch 10, batch 4300, loss[loss=0.2493, simple_loss=0.3298, pruned_loss=0.08445, over 8510.00 frames. ], tot_loss[loss=0.2391, simple_loss=0.3133, pruned_loss=0.08245, over 1616455.53 frames. ], batch size: 26, lr: 7.66e-03, grad_scale: 8.0 +2023-02-06 08:55:52,728 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77052.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:55:55,458 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77056.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 08:55:57,562 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-06 08:56:12,947 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 10, batch 4350, loss[loss=0.2186, simple_loss=0.2884, pruned_loss=0.07436, over 7517.00 frames. ], tot_loss[loss=0.2415, simple_loss=0.3154, pruned_loss=0.08379, over 1617920.08 frames. ], batch size: 18, lr: 7.66e-03, grad_scale: 8.0 +2023-02-06 08:56:24,088 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.7803, 5.9171, 4.9538, 2.4429, 5.0430, 5.4493, 5.3949, 5.1012], + device='cuda:0'), covar=tensor([0.0527, 0.0452, 0.0932, 0.4456, 0.0755, 0.0591, 0.1070, 0.0667], + device='cuda:0'), in_proj_covar=tensor([0.0452, 0.0360, 0.0371, 0.0470, 0.0366, 0.0357, 0.0364, 0.0312], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 08:56:33,987 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77113.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:56:53,837 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-06 08:56:54,996 INFO [optim.py:369] (0/4) 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 08:56:57,034 INFO [train.py:901] (0/4) Epoch 10, batch 4400, loss[loss=0.1847, simple_loss=0.2654, pruned_loss=0.05202, over 7257.00 frames. ], tot_loss[loss=0.2416, simple_loss=0.3151, pruned_loss=0.08406, over 1611131.37 frames. ], batch size: 16, lr: 7.66e-03, grad_scale: 8.0 +2023-02-06 08:57:03,161 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5067, 2.9451, 1.8763, 2.2167, 2.1452, 1.5777, 1.9482, 2.3086], + device='cuda:0'), covar=tensor([0.1324, 0.0275, 0.0927, 0.0599, 0.0656, 0.1279, 0.0972, 0.0865], + device='cuda:0'), in_proj_covar=tensor([0.0338, 0.0230, 0.0304, 0.0290, 0.0297, 0.0317, 0.0333, 0.0305], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 08:57:33,159 INFO [train.py:901] (0/4) Epoch 10, batch 4450, loss[loss=0.2623, simple_loss=0.3372, pruned_loss=0.09374, over 8498.00 frames. ], tot_loss[loss=0.2419, simple_loss=0.3158, pruned_loss=0.08404, over 1610575.81 frames. ], batch size: 26, lr: 7.65e-03, grad_scale: 8.0 +2023-02-06 08:57:35,375 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=77201.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:57:36,013 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-06 08:57:54,223 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 08:58:06,785 INFO [train.py:901] (0/4) Epoch 10, batch 4500, loss[loss=0.2446, simple_loss=0.3243, pruned_loss=0.08242, over 8616.00 frames. ], tot_loss[loss=0.2421, simple_loss=0.3163, pruned_loss=0.08397, over 1612205.23 frames. ], batch size: 39, lr: 7.65e-03, grad_scale: 8.0 +2023-02-06 08:58:27,561 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-06 08:58:36,268 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 +2023-02-06 08:58:38,004 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77291.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:58:41,450 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77295.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:58:43,377 INFO [train.py:901] (0/4) Epoch 10, batch 4550, loss[loss=0.1997, simple_loss=0.2884, pruned_loss=0.05554, over 8139.00 frames. ], tot_loss[loss=0.2416, simple_loss=0.3159, pruned_loss=0.0836, over 1618223.24 frames. ], batch size: 22, lr: 7.65e-03, grad_scale: 8.0 +2023-02-06 08:58:47,524 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.3724, 4.4140, 3.9199, 1.9606, 3.8430, 3.9965, 3.9705, 3.6094], + device='cuda:0'), covar=tensor([0.0742, 0.0589, 0.1039, 0.4684, 0.0792, 0.0898, 0.1427, 0.0731], + device='cuda:0'), in_proj_covar=tensor([0.0454, 0.0358, 0.0372, 0.0471, 0.0370, 0.0357, 0.0369, 0.0311], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 08:58:55,653 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77316.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 08:59:07,702 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4779, 2.8304, 1.8479, 2.1476, 2.2509, 1.4488, 1.9055, 2.1735], + device='cuda:0'), covar=tensor([0.1401, 0.0332, 0.1019, 0.0642, 0.0598, 0.1403, 0.1001, 0.0874], + device='cuda:0'), in_proj_covar=tensor([0.0348, 0.0236, 0.0312, 0.0298, 0.0304, 0.0326, 0.0341, 0.0311], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 08:59:14,841 INFO [optim.py:369] (0/4) 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 08:59:16,946 INFO [train.py:901] (0/4) Epoch 10, batch 4600, loss[loss=0.198, simple_loss=0.2741, pruned_loss=0.06092, over 7415.00 frames. ], tot_loss[loss=0.2428, simple_loss=0.3166, pruned_loss=0.0845, over 1611048.86 frames. ], batch size: 17, lr: 7.65e-03, grad_scale: 8.0 +2023-02-06 08:59:19,226 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8113, 1.9443, 2.1983, 1.5903, 1.2233, 2.3494, 0.2903, 1.4984], + device='cuda:0'), covar=tensor([0.3206, 0.1850, 0.0579, 0.2596, 0.4883, 0.0590, 0.4415, 0.1982], + device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0162, 0.0091, 0.0211, 0.0250, 0.0099, 0.0161, 0.0156], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0001, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 08:59:21,272 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4695, 2.8334, 1.8220, 2.1100, 2.1586, 1.4571, 1.8367, 2.2156], + device='cuda:0'), covar=tensor([0.1460, 0.0298, 0.1131, 0.0697, 0.0734, 0.1475, 0.1075, 0.0946], + device='cuda:0'), in_proj_covar=tensor([0.0349, 0.0237, 0.0315, 0.0299, 0.0305, 0.0327, 0.0342, 0.0312], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 08:59:50,936 INFO [train.py:901] (0/4) Epoch 10, batch 4650, loss[loss=0.2741, simple_loss=0.3416, pruned_loss=0.1033, over 8316.00 frames. ], tot_loss[loss=0.2434, simple_loss=0.3169, pruned_loss=0.08493, over 1612093.88 frames. ], batch size: 26, lr: 7.64e-03, grad_scale: 8.0 +2023-02-06 09:00:19,436 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-02-06 09:00:25,446 INFO [optim.py:369] (0/4) 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 09:00:27,547 INFO [train.py:901] (0/4) Epoch 10, batch 4700, loss[loss=0.2329, simple_loss=0.3143, pruned_loss=0.07574, over 7799.00 frames. ], tot_loss[loss=0.2423, simple_loss=0.316, pruned_loss=0.08434, over 1612797.77 frames. ], batch size: 20, lr: 7.64e-03, grad_scale: 8.0 +2023-02-06 09:00:33,873 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=77457.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:00:59,619 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7731, 1.9524, 2.2654, 1.6203, 1.1779, 2.4452, 0.4435, 1.4491], + device='cuda:0'), covar=tensor([0.2234, 0.1396, 0.0479, 0.2357, 0.4594, 0.0376, 0.3240, 0.1785], + device='cuda:0'), in_proj_covar=tensor([0.0158, 0.0161, 0.0091, 0.0210, 0.0248, 0.0098, 0.0161, 0.0156], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0001, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 09:01:02,738 INFO [train.py:901] (0/4) Epoch 10, batch 4750, loss[loss=0.2465, simple_loss=0.3253, pruned_loss=0.08387, over 8622.00 frames. ], tot_loss[loss=0.2428, simple_loss=0.3165, pruned_loss=0.08455, over 1611215.75 frames. ], batch size: 49, lr: 7.64e-03, grad_scale: 8.0 +2023-02-06 09:01:17,019 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6581, 1.4844, 1.5669, 1.2472, 0.9779, 1.3758, 1.5409, 1.5371], + device='cuda:0'), covar=tensor([0.0534, 0.1214, 0.1725, 0.1442, 0.0571, 0.1529, 0.0679, 0.0581], + device='cuda:0'), in_proj_covar=tensor([0.0101, 0.0155, 0.0195, 0.0160, 0.0105, 0.0165, 0.0118, 0.0139], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0006, 0.0006], + device='cuda:0') +2023-02-06 09:01:25,489 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0863, 1.5783, 1.5723, 1.3616, 1.1432, 1.4486, 1.7606, 1.5973], + device='cuda:0'), covar=tensor([0.0466, 0.1117, 0.1650, 0.1333, 0.0536, 0.1451, 0.0600, 0.0606], + device='cuda:0'), in_proj_covar=tensor([0.0101, 0.0156, 0.0195, 0.0161, 0.0105, 0.0165, 0.0118, 0.0140], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0006, 0.0006], + device='cuda:0') +2023-02-06 09:01:28,493 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-06 09:01:30,513 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-06 09:01:35,870 INFO [optim.py:369] (0/4) 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 09:01:37,926 INFO [train.py:901] (0/4) Epoch 10, batch 4800, loss[loss=0.2555, simple_loss=0.3129, pruned_loss=0.0991, over 7533.00 frames. ], tot_loss[loss=0.2432, simple_loss=0.3167, pruned_loss=0.08483, over 1614048.05 frames. ], batch size: 18, lr: 7.64e-03, grad_scale: 8.0 +2023-02-06 09:01:54,055 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77572.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:01:54,100 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77572.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:01:54,621 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=77573.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:02:10,896 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77597.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:02:11,382 INFO [train.py:901] (0/4) Epoch 10, batch 4850, loss[loss=0.2217, simple_loss=0.2975, pruned_loss=0.07293, over 7926.00 frames. ], tot_loss[loss=0.2438, simple_loss=0.3169, pruned_loss=0.08538, over 1612794.42 frames. ], batch size: 20, lr: 7.63e-03, grad_scale: 8.0 +2023-02-06 09:02:16,304 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-06 09:02:19,184 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.5712, 2.9652, 2.4562, 3.7384, 1.6778, 1.9767, 2.1312, 3.1195], + device='cuda:0'), covar=tensor([0.0696, 0.0787, 0.0965, 0.0320, 0.1265, 0.1416, 0.1313, 0.0645], + device='cuda:0'), in_proj_covar=tensor([0.0239, 0.0215, 0.0260, 0.0221, 0.0224, 0.0253, 0.0263, 0.0223], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 09:02:21,887 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2543, 2.1920, 1.6813, 1.9709, 1.8163, 1.2910, 1.5109, 1.6478], + device='cuda:0'), covar=tensor([0.1198, 0.0332, 0.1000, 0.0446, 0.0564, 0.1459, 0.0923, 0.0779], + device='cuda:0'), in_proj_covar=tensor([0.0348, 0.0236, 0.0314, 0.0300, 0.0303, 0.0328, 0.0342, 0.0311], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 09:02:29,382 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.6581, 1.7200, 4.9124, 1.8040, 4.3174, 4.1946, 4.4318, 4.2790], + device='cuda:0'), covar=tensor([0.0429, 0.3694, 0.0390, 0.3205, 0.1103, 0.0873, 0.0428, 0.0571], + device='cuda:0'), in_proj_covar=tensor([0.0457, 0.0558, 0.0562, 0.0517, 0.0591, 0.0501, 0.0495, 0.0566], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 09:02:37,996 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77634.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:02:38,613 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=77635.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:02:41,995 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=77639.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:02:46,052 INFO [optim.py:369] (0/4) 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 09:02:48,054 INFO [train.py:901] (0/4) Epoch 10, batch 4900, loss[loss=0.2813, simple_loss=0.346, pruned_loss=0.1083, over 8468.00 frames. ], tot_loss[loss=0.2437, simple_loss=0.317, pruned_loss=0.08515, over 1612074.90 frames. ], batch size: 25, lr: 7.63e-03, grad_scale: 8.0 +2023-02-06 09:02:51,344 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 +2023-02-06 09:03:08,153 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.57 vs. limit=2.0 +2023-02-06 09:03:15,419 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77688.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:03:18,975 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 +2023-02-06 09:03:19,985 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77695.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:03:21,734 INFO [train.py:901] (0/4) Epoch 10, batch 4950, loss[loss=0.2525, simple_loss=0.3325, pruned_loss=0.08626, over 8335.00 frames. ], tot_loss[loss=0.2445, simple_loss=0.3177, pruned_loss=0.08572, over 1616496.29 frames. ], batch size: 26, lr: 7.63e-03, grad_scale: 8.0 +2023-02-06 09:03:26,726 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.71 vs. limit=5.0 +2023-02-06 09:03:54,524 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 10, batch 5000, loss[loss=0.2289, simple_loss=0.2904, pruned_loss=0.08373, over 7432.00 frames. ], tot_loss[loss=0.2435, simple_loss=0.3168, pruned_loss=0.08513, over 1613228.37 frames. ], batch size: 17, lr: 7.63e-03, grad_scale: 8.0 +2023-02-06 09:03:58,685 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77750.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:04:01,926 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77754.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:04:10,213 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 +2023-02-06 09:04:19,116 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.7709, 2.1760, 3.6368, 2.7859, 3.1529, 2.4877, 1.9089, 1.8801], + device='cuda:0'), covar=tensor([0.3190, 0.3984, 0.1002, 0.2379, 0.1773, 0.1940, 0.1637, 0.4283], + device='cuda:0'), in_proj_covar=tensor([0.0879, 0.0851, 0.0717, 0.0829, 0.0927, 0.0782, 0.0699, 0.0757], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 09:04:30,717 INFO [train.py:901] (0/4) Epoch 10, batch 5050, loss[loss=0.27, simple_loss=0.3299, pruned_loss=0.1051, over 8131.00 frames. ], tot_loss[loss=0.2439, simple_loss=0.3174, pruned_loss=0.08525, over 1618901.31 frames. ], batch size: 22, lr: 7.62e-03, grad_scale: 8.0 +2023-02-06 09:04:51,126 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77828.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:04:52,878 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-06 09:05:02,962 INFO [optim.py:369] (0/4) 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 09:05:05,647 INFO [train.py:901] (0/4) Epoch 10, batch 5100, loss[loss=0.2621, simple_loss=0.3137, pruned_loss=0.1052, over 7515.00 frames. ], tot_loss[loss=0.2413, simple_loss=0.3154, pruned_loss=0.08356, over 1620091.96 frames. ], batch size: 18, lr: 7.62e-03, grad_scale: 16.0 +2023-02-06 09:05:09,146 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77853.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:05:28,990 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9570, 2.3904, 2.6363, 1.5173, 2.8152, 1.7792, 1.5737, 1.8809], + device='cuda:0'), covar=tensor([0.0411, 0.0230, 0.0149, 0.0393, 0.0209, 0.0429, 0.0446, 0.0264], + device='cuda:0'), in_proj_covar=tensor([0.0370, 0.0306, 0.0261, 0.0369, 0.0296, 0.0454, 0.0342, 0.0332], + device='cuda:0'), out_proj_covar=tensor([1.0746e-04, 8.6909e-05, 7.4505e-05, 1.0521e-04, 8.5564e-05, 1.4086e-04, + 9.9553e-05, 9.5814e-05], device='cuda:0') +2023-02-06 09:05:40,070 INFO [train.py:901] (0/4) Epoch 10, batch 5150, loss[loss=0.203, simple_loss=0.2857, pruned_loss=0.06014, over 8196.00 frames. ], tot_loss[loss=0.241, simple_loss=0.3152, pruned_loss=0.08342, over 1621605.00 frames. ], batch size: 23, lr: 7.62e-03, grad_scale: 16.0 +2023-02-06 09:06:10,562 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7466, 1.9080, 2.2559, 1.7169, 1.3658, 2.2657, 0.3712, 1.3895], + device='cuda:0'), covar=tensor([0.2727, 0.1621, 0.0516, 0.1835, 0.4169, 0.0529, 0.3907, 0.2194], + device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0164, 0.0092, 0.0211, 0.0250, 0.0098, 0.0161, 0.0161], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0001, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 09:06:11,264 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77944.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:06:11,731 INFO [optim.py:369] (0/4) 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 09:06:13,804 INFO [train.py:901] (0/4) Epoch 10, batch 5200, loss[loss=0.2117, simple_loss=0.2804, pruned_loss=0.07146, over 6773.00 frames. ], tot_loss[loss=0.2403, simple_loss=0.3144, pruned_loss=0.08315, over 1615353.74 frames. ], batch size: 15, lr: 7.62e-03, grad_scale: 16.0 +2023-02-06 09:06:29,068 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77969.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:06:35,807 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=77978.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:06:50,964 INFO [train.py:901] (0/4) Epoch 10, batch 5250, loss[loss=0.1948, simple_loss=0.2836, pruned_loss=0.05297, over 8133.00 frames. ], tot_loss[loss=0.2401, simple_loss=0.3142, pruned_loss=0.08301, over 1617678.98 frames. ], batch size: 22, lr: 7.61e-03, grad_scale: 16.0 +2023-02-06 09:06:52,496 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-78000.pt +2023-02-06 09:06:56,852 WARNING [train.py:1067] (0/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-06 09:06:57,791 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=78006.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:07:00,626 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=78010.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:07:14,846 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78031.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:07:17,393 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78035.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:07:19,991 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 09:07:25,980 INFO [train.py:901] (0/4) Epoch 10, batch 5300, loss[loss=0.2468, simple_loss=0.3245, pruned_loss=0.08458, over 8498.00 frames. ], tot_loss[loss=0.2397, simple_loss=0.3138, pruned_loss=0.08277, over 1623159.30 frames. ], batch size: 28, lr: 7.61e-03, grad_scale: 16.0 +2023-02-06 09:07:57,634 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=78093.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:08:00,802 INFO [train.py:901] (0/4) Epoch 10, batch 5350, loss[loss=0.281, simple_loss=0.3479, pruned_loss=0.1071, over 8744.00 frames. ], tot_loss[loss=0.2405, simple_loss=0.3142, pruned_loss=0.08338, over 1622460.52 frames. ], batch size: 30, lr: 7.61e-03, grad_scale: 16.0 +2023-02-06 09:08:34,365 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 10, batch 5400, loss[loss=0.2665, simple_loss=0.3382, pruned_loss=0.09734, over 8551.00 frames. ], tot_loss[loss=0.2409, simple_loss=0.3143, pruned_loss=0.08372, over 1618540.99 frames. ], batch size: 31, lr: 7.61e-03, grad_scale: 8.0 +2023-02-06 09:08:40,001 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=78154.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:09:07,680 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.58 vs. limit=5.0 +2023-02-06 09:09:08,707 INFO [train.py:901] (0/4) Epoch 10, batch 5450, loss[loss=0.2177, simple_loss=0.2908, pruned_loss=0.07233, over 7925.00 frames. ], tot_loss[loss=0.2416, simple_loss=0.3148, pruned_loss=0.08424, over 1618059.85 frames. ], batch size: 20, lr: 7.60e-03, grad_scale: 8.0 +2023-02-06 09:09:11,046 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.2722, 1.2541, 4.4886, 1.6963, 3.8560, 3.6198, 3.9811, 3.8266], + device='cuda:0'), covar=tensor([0.0546, 0.4394, 0.0496, 0.3503, 0.1252, 0.0915, 0.0595, 0.0641], + device='cuda:0'), in_proj_covar=tensor([0.0463, 0.0560, 0.0562, 0.0520, 0.0600, 0.0503, 0.0497, 0.0566], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 09:09:17,777 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 +2023-02-06 09:09:25,579 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9311, 2.1904, 3.3523, 1.6680, 2.6744, 2.2876, 2.0550, 2.6285], + device='cuda:0'), covar=tensor([0.1260, 0.1815, 0.0632, 0.2944, 0.1214, 0.2038, 0.1334, 0.1589], + device='cuda:0'), in_proj_covar=tensor([0.0486, 0.0508, 0.0531, 0.0575, 0.0609, 0.0551, 0.0466, 0.0606], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 09:09:43,438 INFO [optim.py:369] (0/4) 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 09:09:44,805 INFO [train.py:901] (0/4) Epoch 10, batch 5500, loss[loss=0.2311, simple_loss=0.3222, pruned_loss=0.07002, over 8251.00 frames. ], tot_loss[loss=0.2407, simple_loss=0.3142, pruned_loss=0.08358, over 1613720.75 frames. ], batch size: 24, lr: 7.60e-03, grad_scale: 8.0 +2023-02-06 09:09:46,323 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78250.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:09:46,834 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-06 09:10:18,555 INFO [train.py:901] (0/4) Epoch 10, batch 5550, loss[loss=0.2803, simple_loss=0.3452, pruned_loss=0.1077, over 8257.00 frames. ], tot_loss[loss=0.2424, simple_loss=0.3159, pruned_loss=0.08439, over 1612255.35 frames. ], batch size: 24, lr: 7.60e-03, grad_scale: 8.0 +2023-02-06 09:10:32,884 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 +2023-02-06 09:10:42,880 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.35 vs. limit=5.0 +2023-02-06 09:10:52,577 INFO [optim.py:369] (0/4) 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 09:10:54,679 INFO [train.py:901] (0/4) Epoch 10, batch 5600, loss[loss=0.2201, simple_loss=0.2947, pruned_loss=0.07276, over 8293.00 frames. ], tot_loss[loss=0.2412, simple_loss=0.3149, pruned_loss=0.08378, over 1613151.59 frames. ], batch size: 23, lr: 7.60e-03, grad_scale: 8.0 +2023-02-06 09:10:55,585 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=78349.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:11:00,335 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5083, 1.8126, 1.8513, 1.2978, 2.0311, 1.4021, 0.4635, 1.7315], + device='cuda:0'), covar=tensor([0.0310, 0.0200, 0.0175, 0.0268, 0.0218, 0.0565, 0.0487, 0.0143], + device='cuda:0'), in_proj_covar=tensor([0.0374, 0.0307, 0.0263, 0.0372, 0.0300, 0.0459, 0.0347, 0.0332], + device='cuda:0'), out_proj_covar=tensor([1.0823e-04, 8.6926e-05, 7.4862e-05, 1.0610e-04, 8.6683e-05, 1.4272e-04, + 1.0089e-04, 9.5404e-05], device='cuda:0') +2023-02-06 09:11:12,347 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78374.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:11:15,171 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.45 vs. limit=5.0 +2023-02-06 09:11:28,359 INFO [train.py:901] (0/4) Epoch 10, batch 5650, loss[loss=0.199, simple_loss=0.2886, pruned_loss=0.0547, over 7649.00 frames. ], tot_loss[loss=0.2407, simple_loss=0.3147, pruned_loss=0.08336, over 1613272.52 frames. ], batch size: 19, lr: 7.59e-03, grad_scale: 8.0 +2023-02-06 09:11:28,796 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.52 vs. limit=5.0 +2023-02-06 09:11:36,926 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=78410.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:11:48,309 WARNING [train.py:1067] (0/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-06 09:11:54,453 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78435.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:12:02,397 INFO [optim.py:369] (0/4) 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 09:12:03,745 INFO [train.py:901] (0/4) Epoch 10, batch 5700, loss[loss=0.236, simple_loss=0.3043, pruned_loss=0.08385, over 7650.00 frames. ], tot_loss[loss=0.241, simple_loss=0.3154, pruned_loss=0.0833, over 1612003.72 frames. ], batch size: 19, lr: 7.59e-03, grad_scale: 8.0 +2023-02-06 09:12:19,552 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7148, 1.7388, 2.0276, 1.6189, 1.2605, 2.0275, 0.2612, 1.2887], + device='cuda:0'), covar=tensor([0.2623, 0.1796, 0.0510, 0.1675, 0.4255, 0.0666, 0.3458, 0.2003], + device='cuda:0'), in_proj_covar=tensor([0.0159, 0.0163, 0.0091, 0.0208, 0.0249, 0.0098, 0.0159, 0.0158], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0001, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 09:12:25,605 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78478.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:12:26,994 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0818, 1.5458, 3.4527, 1.4744, 2.2177, 3.8168, 3.8131, 3.2782], + device='cuda:0'), covar=tensor([0.0862, 0.1451, 0.0303, 0.1740, 0.0911, 0.0197, 0.0420, 0.0565], + device='cuda:0'), in_proj_covar=tensor([0.0254, 0.0291, 0.0252, 0.0282, 0.0266, 0.0230, 0.0322, 0.0283], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:0') +2023-02-06 09:12:27,752 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4954, 1.9286, 3.6392, 1.2879, 2.4380, 1.9701, 1.6017, 2.3156], + device='cuda:0'), covar=tensor([0.1728, 0.2147, 0.0606, 0.3729, 0.1656, 0.2731, 0.1728, 0.2185], + device='cuda:0'), in_proj_covar=tensor([0.0485, 0.0506, 0.0529, 0.0573, 0.0613, 0.0550, 0.0465, 0.0604], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 09:12:39,142 INFO [train.py:901] (0/4) Epoch 10, batch 5750, loss[loss=0.1736, simple_loss=0.2499, pruned_loss=0.04865, over 7660.00 frames. ], tot_loss[loss=0.2412, simple_loss=0.3155, pruned_loss=0.08345, over 1610103.52 frames. ], batch size: 19, lr: 7.59e-03, grad_scale: 8.0 +2023-02-06 09:12:53,272 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-06 09:13:11,243 INFO [optim.py:369] (0/4) 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 09:13:12,605 INFO [train.py:901] (0/4) Epoch 10, batch 5800, loss[loss=0.2582, simple_loss=0.3408, pruned_loss=0.08781, over 8473.00 frames. ], tot_loss[loss=0.2416, simple_loss=0.3157, pruned_loss=0.08373, over 1609635.69 frames. ], batch size: 25, lr: 7.59e-03, grad_scale: 8.0 +2023-02-06 09:13:31,624 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78574.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:13:45,411 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=78594.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:13:48,015 INFO [train.py:901] (0/4) Epoch 10, batch 5850, loss[loss=0.2992, simple_loss=0.3615, pruned_loss=0.1185, over 8337.00 frames. ], tot_loss[loss=0.2427, simple_loss=0.3164, pruned_loss=0.08455, over 1609721.11 frames. ], batch size: 25, lr: 7.58e-03, grad_scale: 8.0 +2023-02-06 09:14:19,894 INFO [optim.py:369] (0/4) 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 09:14:21,258 INFO [train.py:901] (0/4) Epoch 10, batch 5900, loss[loss=0.246, simple_loss=0.3233, pruned_loss=0.08432, over 8501.00 frames. ], tot_loss[loss=0.2409, simple_loss=0.315, pruned_loss=0.08334, over 1610518.46 frames. ], batch size: 29, lr: 7.58e-03, grad_scale: 8.0 +2023-02-06 09:14:32,172 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6037, 1.3932, 1.5912, 1.2334, 0.8966, 1.3787, 1.6487, 1.6111], + device='cuda:0'), covar=tensor([0.0577, 0.1295, 0.1824, 0.1401, 0.0601, 0.1615, 0.0684, 0.0612], + device='cuda:0'), in_proj_covar=tensor([0.0103, 0.0156, 0.0196, 0.0160, 0.0106, 0.0166, 0.0119, 0.0141], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0006], + device='cuda:0') +2023-02-06 09:14:57,568 INFO [train.py:901] (0/4) Epoch 10, batch 5950, loss[loss=0.1959, simple_loss=0.2715, pruned_loss=0.0602, over 7792.00 frames. ], tot_loss[loss=0.2396, simple_loss=0.3136, pruned_loss=0.08282, over 1610967.82 frames. ], batch size: 19, lr: 7.58e-03, grad_scale: 8.0 +2023-02-06 09:15:05,417 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=78709.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:15:30,047 INFO [optim.py:369] (0/4) 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 09:15:31,435 INFO [train.py:901] (0/4) Epoch 10, batch 6000, loss[loss=0.2505, simple_loss=0.3363, pruned_loss=0.08232, over 8491.00 frames. ], tot_loss[loss=0.2408, simple_loss=0.3147, pruned_loss=0.08347, over 1614859.61 frames. ], batch size: 29, lr: 7.58e-03, grad_scale: 8.0 +2023-02-06 09:15:31,436 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 09:15:43,955 INFO [train.py:935] (0/4) Epoch 10, validation: loss=0.1914, simple_loss=0.2907, pruned_loss=0.04604, over 944034.00 frames. +2023-02-06 09:15:43,956 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-06 09:16:14,654 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=2.12 vs. limit=2.0 +2023-02-06 09:16:18,420 INFO [train.py:901] (0/4) Epoch 10, batch 6050, loss[loss=0.2638, simple_loss=0.3281, pruned_loss=0.09973, over 7403.00 frames. ], tot_loss[loss=0.2408, simple_loss=0.3146, pruned_loss=0.08346, over 1612616.78 frames. ], batch size: 72, lr: 7.58e-03, grad_scale: 8.0 +2023-02-06 09:16:35,958 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=78822.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:16:40,303 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 +2023-02-06 09:16:52,859 INFO [optim.py:369] (0/4) 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 09:16:54,181 INFO [train.py:901] (0/4) Epoch 10, batch 6100, loss[loss=0.2304, simple_loss=0.3144, pruned_loss=0.07316, over 8077.00 frames. ], tot_loss[loss=0.2422, simple_loss=0.3157, pruned_loss=0.08438, over 1616293.37 frames. ], batch size: 21, lr: 7.57e-03, grad_scale: 8.0 +2023-02-06 09:17:24,368 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 10, batch 6150, loss[loss=0.2552, simple_loss=0.3367, pruned_loss=0.08684, over 8332.00 frames. ], tot_loss[loss=0.2417, simple_loss=0.3155, pruned_loss=0.08394, over 1619540.80 frames. ], batch size: 25, lr: 7.57e-03, grad_scale: 8.0 +2023-02-06 09:17:41,311 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=78918.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:17:51,534 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 +2023-02-06 09:17:54,625 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=78937.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:18:01,039 INFO [optim.py:369] (0/4) 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 09:18:02,472 INFO [train.py:901] (0/4) Epoch 10, batch 6200, loss[loss=0.2386, simple_loss=0.3168, pruned_loss=0.08023, over 8690.00 frames. ], tot_loss[loss=0.2413, simple_loss=0.3152, pruned_loss=0.08364, over 1622002.85 frames. ], batch size: 34, lr: 7.57e-03, grad_scale: 8.0 +2023-02-06 09:18:15,684 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=78965.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:18:28,060 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78983.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:18:32,919 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78990.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:18:38,071 INFO [train.py:901] (0/4) Epoch 10, batch 6250, loss[loss=0.197, simple_loss=0.2712, pruned_loss=0.06135, over 7700.00 frames. ], tot_loss[loss=0.2403, simple_loss=0.3144, pruned_loss=0.08309, over 1622500.26 frames. ], batch size: 18, lr: 7.57e-03, grad_scale: 8.0 +2023-02-06 09:18:47,162 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3586, 1.9598, 3.1994, 1.1837, 2.3286, 1.7598, 1.4824, 2.1293], + device='cuda:0'), covar=tensor([0.1813, 0.2083, 0.0737, 0.3850, 0.1621, 0.2969, 0.1819, 0.2161], + device='cuda:0'), in_proj_covar=tensor([0.0486, 0.0511, 0.0533, 0.0576, 0.0613, 0.0554, 0.0467, 0.0606], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 09:19:01,795 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79033.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:19:07,232 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-02-06 09:19:10,142 INFO [optim.py:369] (0/4) 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 09:19:11,555 INFO [train.py:901] (0/4) Epoch 10, batch 6300, loss[loss=0.245, simple_loss=0.3314, pruned_loss=0.0793, over 8403.00 frames. ], tot_loss[loss=0.2409, simple_loss=0.3149, pruned_loss=0.08347, over 1620079.53 frames. ], batch size: 49, lr: 7.56e-03, grad_scale: 8.0 +2023-02-06 09:19:47,655 INFO [train.py:901] (0/4) Epoch 10, batch 6350, loss[loss=0.2628, simple_loss=0.3139, pruned_loss=0.1059, over 8336.00 frames. ], tot_loss[loss=0.2423, simple_loss=0.3153, pruned_loss=0.08465, over 1615389.38 frames. ], batch size: 26, lr: 7.56e-03, grad_scale: 8.0 +2023-02-06 09:20:00,245 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-06 09:20:01,435 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2209, 1.8512, 2.7599, 2.2397, 2.4300, 2.0257, 1.5843, 1.2339], + device='cuda:0'), covar=tensor([0.3828, 0.3837, 0.1023, 0.2029, 0.1662, 0.2126, 0.1954, 0.3410], + device='cuda:0'), in_proj_covar=tensor([0.0865, 0.0842, 0.0701, 0.0816, 0.0911, 0.0772, 0.0685, 0.0742], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 09:20:20,614 INFO [optim.py:369] (0/4) 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 09:20:21,297 INFO [train.py:901] (0/4) Epoch 10, batch 6400, loss[loss=0.2925, simple_loss=0.3481, pruned_loss=0.1185, over 8331.00 frames. ], tot_loss[loss=0.2421, simple_loss=0.3153, pruned_loss=0.08445, over 1617246.06 frames. ], batch size: 25, lr: 7.56e-03, grad_scale: 8.0 +2023-02-06 09:20:54,142 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79193.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:20:57,426 INFO [train.py:901] (0/4) Epoch 10, batch 6450, loss[loss=0.1743, simple_loss=0.2552, pruned_loss=0.04668, over 7717.00 frames. ], tot_loss[loss=0.2408, simple_loss=0.3141, pruned_loss=0.08375, over 1614056.67 frames. ], batch size: 18, lr: 7.56e-03, grad_scale: 8.0 +2023-02-06 09:21:12,196 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79218.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:21:27,157 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1544, 1.7818, 2.5390, 1.9840, 2.2599, 1.9835, 1.6358, 0.9979], + device='cuda:0'), covar=tensor([0.3553, 0.3448, 0.1109, 0.2457, 0.1856, 0.2125, 0.1729, 0.3900], + device='cuda:0'), in_proj_covar=tensor([0.0871, 0.0846, 0.0707, 0.0821, 0.0918, 0.0775, 0.0690, 0.0748], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 09:21:31,632 INFO [optim.py:369] (0/4) 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 09:21:32,337 INFO [train.py:901] (0/4) Epoch 10, batch 6500, loss[loss=0.2508, simple_loss=0.3261, pruned_loss=0.08772, over 8166.00 frames. ], tot_loss[loss=0.2409, simple_loss=0.3148, pruned_loss=0.08348, over 1614657.35 frames. ], batch size: 48, lr: 7.55e-03, grad_scale: 8.0 +2023-02-06 09:21:35,280 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 +2023-02-06 09:21:49,240 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3697, 1.8688, 2.9102, 2.1912, 2.6547, 2.0709, 1.6925, 1.2402], + device='cuda:0'), covar=tensor([0.3634, 0.3798, 0.1034, 0.2579, 0.1721, 0.2056, 0.1714, 0.4142], + device='cuda:0'), in_proj_covar=tensor([0.0880, 0.0853, 0.0714, 0.0830, 0.0925, 0.0782, 0.0699, 0.0756], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 09:21:59,867 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 10, batch 6550, loss[loss=0.3027, simple_loss=0.3651, pruned_loss=0.1201, over 6527.00 frames. ], tot_loss[loss=0.2401, simple_loss=0.314, pruned_loss=0.08304, over 1611596.75 frames. ], batch size: 71, lr: 7.55e-03, grad_scale: 8.0 +2023-02-06 09:22:11,149 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79303.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 09:22:18,365 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=79327.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:22:36,563 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-06 09:22:41,242 INFO [optim.py:369] (0/4) 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 09:22:41,948 INFO [train.py:901] (0/4) Epoch 10, batch 6600, loss[loss=0.2415, simple_loss=0.3264, pruned_loss=0.07834, over 8254.00 frames. ], tot_loss[loss=0.2412, simple_loss=0.3151, pruned_loss=0.08363, over 1613711.99 frames. ], batch size: 24, lr: 7.55e-03, grad_scale: 8.0 +2023-02-06 09:22:44,149 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79351.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:22:53,884 WARNING [train.py:1067] (0/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-06 09:23:04,713 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79382.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:23:11,888 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79393.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 09:23:15,196 INFO [train.py:901] (0/4) Epoch 10, batch 6650, loss[loss=0.2396, simple_loss=0.3118, pruned_loss=0.08375, over 8532.00 frames. ], tot_loss[loss=0.2438, simple_loss=0.3172, pruned_loss=0.08514, over 1617970.96 frames. ], batch size: 28, lr: 7.55e-03, grad_scale: 8.0 +2023-02-06 09:23:33,628 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4747, 2.9202, 1.8714, 2.3958, 2.3115, 1.5585, 2.1337, 2.2231], + device='cuda:0'), covar=tensor([0.1437, 0.0292, 0.1009, 0.0581, 0.0568, 0.1295, 0.0891, 0.0902], + device='cuda:0'), in_proj_covar=tensor([0.0344, 0.0229, 0.0308, 0.0296, 0.0303, 0.0319, 0.0336, 0.0305], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 09:23:47,672 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79442.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:23:50,867 INFO [optim.py:369] (0/4) 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 09:23:51,581 INFO [train.py:901] (0/4) Epoch 10, batch 6700, loss[loss=0.2776, simple_loss=0.3415, pruned_loss=0.1069, over 8135.00 frames. ], tot_loss[loss=0.2428, simple_loss=0.3162, pruned_loss=0.08464, over 1614299.56 frames. ], batch size: 22, lr: 7.54e-03, grad_scale: 8.0 +2023-02-06 09:24:24,670 INFO [train.py:901] (0/4) Epoch 10, batch 6750, loss[loss=0.2307, simple_loss=0.2988, pruned_loss=0.08128, over 7671.00 frames. ], tot_loss[loss=0.2444, simple_loss=0.3174, pruned_loss=0.08574, over 1615830.19 frames. ], batch size: 19, lr: 7.54e-03, grad_scale: 8.0 +2023-02-06 09:24:28,996 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1133, 2.7610, 2.2343, 2.3861, 2.3943, 2.0185, 2.2220, 2.4415], + device='cuda:0'), covar=tensor([0.0885, 0.0221, 0.0678, 0.0464, 0.0468, 0.0892, 0.0682, 0.0645], + device='cuda:0'), in_proj_covar=tensor([0.0345, 0.0229, 0.0308, 0.0296, 0.0305, 0.0320, 0.0338, 0.0306], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 09:25:00,368 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 10, batch 6800, loss[loss=0.1741, simple_loss=0.2599, pruned_loss=0.04417, over 7422.00 frames. ], tot_loss[loss=0.2422, simple_loss=0.3156, pruned_loss=0.08442, over 1614639.23 frames. ], batch size: 17, lr: 7.54e-03, grad_scale: 8.0 +2023-02-06 09:25:11,659 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-06 09:25:14,138 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 +2023-02-06 09:25:19,537 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4186, 1.4610, 1.3622, 1.8805, 0.7612, 1.1813, 1.3369, 1.4906], + device='cuda:0'), covar=tensor([0.0832, 0.0828, 0.1185, 0.0565, 0.1179, 0.1513, 0.0805, 0.0782], + device='cuda:0'), in_proj_covar=tensor([0.0239, 0.0211, 0.0254, 0.0215, 0.0218, 0.0251, 0.0258, 0.0224], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 09:25:35,817 INFO [train.py:901] (0/4) Epoch 10, batch 6850, loss[loss=0.2493, simple_loss=0.3311, pruned_loss=0.08377, over 8359.00 frames. ], tot_loss[loss=0.2409, simple_loss=0.3147, pruned_loss=0.08357, over 1614852.29 frames. ], batch size: 24, lr: 7.54e-03, grad_scale: 8.0 +2023-02-06 09:25:39,343 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79603.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:25:59,645 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-06 09:26:10,489 INFO [optim.py:369] (0/4) 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 09:26:10,574 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=79647.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 09:26:11,113 INFO [train.py:901] (0/4) Epoch 10, batch 6900, loss[loss=0.2567, simple_loss=0.3309, pruned_loss=0.09122, over 8502.00 frames. ], tot_loss[loss=0.2396, simple_loss=0.3138, pruned_loss=0.08268, over 1614813.99 frames. ], batch size: 26, lr: 7.53e-03, grad_scale: 8.0 +2023-02-06 09:26:30,977 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79675.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:26:44,397 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=79695.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:26:46,400 INFO [train.py:901] (0/4) Epoch 10, batch 6950, loss[loss=0.2395, simple_loss=0.3088, pruned_loss=0.08512, over 7804.00 frames. ], tot_loss[loss=0.2399, simple_loss=0.3144, pruned_loss=0.08269, over 1615173.68 frames. ], batch size: 20, lr: 7.53e-03, grad_scale: 8.0 +2023-02-06 09:26:46,618 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79698.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:27:03,548 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79723.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:27:05,482 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=79726.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:27:10,656 WARNING [train.py:1067] (0/4) 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] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=79737.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 09:27:19,257 INFO [optim.py:369] (0/4) 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 09:27:19,980 INFO [train.py:901] (0/4) Epoch 10, batch 7000, loss[loss=0.2748, simple_loss=0.3375, pruned_loss=0.106, over 8789.00 frames. ], tot_loss[loss=0.241, simple_loss=0.3147, pruned_loss=0.08367, over 1612165.11 frames. ], batch size: 30, lr: 7.53e-03, grad_scale: 8.0 +2023-02-06 09:27:30,359 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79762.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 09:27:43,435 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79780.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:27:55,331 INFO [train.py:901] (0/4) Epoch 10, batch 7050, loss[loss=0.2159, simple_loss=0.2793, pruned_loss=0.07628, over 7711.00 frames. ], tot_loss[loss=0.2406, simple_loss=0.3142, pruned_loss=0.08349, over 1609374.94 frames. ], batch size: 18, lr: 7.53e-03, grad_scale: 8.0 +2023-02-06 09:28:04,452 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79810.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:28:25,524 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79841.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:28:26,876 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.2036, 1.7050, 1.9344, 1.6083, 1.3493, 1.8419, 2.4731, 2.7678], + device='cuda:0'), covar=tensor([0.0435, 0.1229, 0.1636, 0.1352, 0.0601, 0.1454, 0.0596, 0.0428], + device='cuda:0'), in_proj_covar=tensor([0.0102, 0.0152, 0.0194, 0.0159, 0.0106, 0.0164, 0.0118, 0.0138], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0006], + device='cuda:0') +2023-02-06 09:28:29,361 INFO [optim.py:369] (0/4) 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 09:28:30,076 INFO [train.py:901] (0/4) Epoch 10, batch 7100, loss[loss=0.2482, simple_loss=0.3194, pruned_loss=0.08854, over 8550.00 frames. ], tot_loss[loss=0.2405, simple_loss=0.3142, pruned_loss=0.08344, over 1609969.81 frames. ], batch size: 31, lr: 7.53e-03, grad_scale: 8.0 +2023-02-06 09:28:32,878 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79852.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 09:29:06,027 INFO [train.py:901] (0/4) Epoch 10, batch 7150, loss[loss=0.2623, simple_loss=0.3411, pruned_loss=0.09182, over 8188.00 frames. ], tot_loss[loss=0.2413, simple_loss=0.3152, pruned_loss=0.0837, over 1608338.76 frames. ], batch size: 23, lr: 7.52e-03, grad_scale: 8.0 +2023-02-06 09:29:16,898 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3479, 1.2778, 1.5729, 1.2493, 0.7722, 1.3742, 1.1978, 1.3112], + device='cuda:0'), covar=tensor([0.0556, 0.1182, 0.1758, 0.1396, 0.0594, 0.1479, 0.0672, 0.0607], + device='cuda:0'), in_proj_covar=tensor([0.0102, 0.0153, 0.0195, 0.0160, 0.0105, 0.0164, 0.0118, 0.0139], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0006], + device='cuda:0') +2023-02-06 09:29:30,376 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9753, 1.6086, 2.1646, 1.8048, 1.9857, 1.9059, 1.5827, 0.7014], + device='cuda:0'), covar=tensor([0.3923, 0.3643, 0.1201, 0.2191, 0.1754, 0.1966, 0.1660, 0.3752], + device='cuda:0'), in_proj_covar=tensor([0.0877, 0.0852, 0.0716, 0.0823, 0.0925, 0.0783, 0.0698, 0.0759], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 09:29:39,480 INFO [optim.py:369] (0/4) 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 09:29:39,570 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=79947.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:29:40,183 INFO [train.py:901] (0/4) Epoch 10, batch 7200, loss[loss=0.2382, simple_loss=0.2987, pruned_loss=0.08888, over 7792.00 frames. ], tot_loss[loss=0.2419, simple_loss=0.3155, pruned_loss=0.08415, over 1611302.70 frames. ], batch size: 19, lr: 7.52e-03, grad_scale: 8.0 +2023-02-06 09:30:09,195 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.4777, 1.3545, 4.6828, 1.7735, 4.0945, 3.8708, 4.1848, 4.0596], + device='cuda:0'), covar=tensor([0.0543, 0.4201, 0.0436, 0.3215, 0.1102, 0.0754, 0.0517, 0.0616], + device='cuda:0'), in_proj_covar=tensor([0.0464, 0.0560, 0.0564, 0.0519, 0.0587, 0.0498, 0.0491, 0.0562], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 09:30:13,865 INFO [train.py:901] (0/4) Epoch 10, batch 7250, loss[loss=0.207, simple_loss=0.2784, pruned_loss=0.06782, over 7211.00 frames. ], tot_loss[loss=0.2394, simple_loss=0.3131, pruned_loss=0.08285, over 1608868.74 frames. ], batch size: 16, lr: 7.52e-03, grad_scale: 8.0 +2023-02-06 09:30:14,881 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 +2023-02-06 09:30:15,334 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-80000.pt +2023-02-06 09:30:30,524 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80018.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 09:30:31,020 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=80019.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:30:33,692 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6566, 1.8463, 2.2307, 1.7798, 1.1909, 2.3239, 0.3295, 1.4778], + device='cuda:0'), covar=tensor([0.3276, 0.1661, 0.0463, 0.2047, 0.4433, 0.0521, 0.3468, 0.1918], + device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0163, 0.0093, 0.0208, 0.0253, 0.0100, 0.0159, 0.0160], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0001, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 09:30:47,886 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) Epoch 10, batch 7300, loss[loss=0.1869, simple_loss=0.2609, pruned_loss=0.05645, over 7441.00 frames. ], tot_loss[loss=0.239, simple_loss=0.3128, pruned_loss=0.08263, over 1607185.15 frames. ], batch size: 17, lr: 7.52e-03, grad_scale: 8.0 +2023-02-06 09:31:00,590 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80062.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:31:03,340 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80066.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:31:04,626 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.1133, 4.1228, 3.7510, 2.1288, 3.6787, 3.7306, 3.7788, 3.4577], + device='cuda:0'), covar=tensor([0.0905, 0.0596, 0.1022, 0.4064, 0.0928, 0.0946, 0.1321, 0.0767], + device='cuda:0'), in_proj_covar=tensor([0.0450, 0.0354, 0.0370, 0.0466, 0.0369, 0.0354, 0.0361, 0.0313], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 09:31:19,981 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80091.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:31:24,176 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80097.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:31:24,603 INFO [train.py:901] (0/4) Epoch 10, batch 7350, loss[loss=0.2683, simple_loss=0.3318, pruned_loss=0.1024, over 8474.00 frames. ], tot_loss[loss=0.2405, simple_loss=0.3146, pruned_loss=0.08316, over 1610551.93 frames. ], batch size: 25, lr: 7.51e-03, grad_scale: 8.0 +2023-02-06 09:31:27,536 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0649, 2.2892, 1.7979, 2.7327, 1.2135, 1.5026, 1.6734, 2.3206], + device='cuda:0'), covar=tensor([0.0802, 0.0877, 0.1107, 0.0457, 0.1328, 0.1583, 0.1237, 0.0796], + device='cuda:0'), in_proj_covar=tensor([0.0240, 0.0214, 0.0254, 0.0218, 0.0219, 0.0251, 0.0259, 0.0227], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 09:31:31,669 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80108.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 09:31:32,372 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0769, 1.7642, 2.5718, 2.0437, 2.3227, 2.0211, 1.6514, 0.9999], + device='cuda:0'), covar=tensor([0.4027, 0.3843, 0.1135, 0.2329, 0.1747, 0.2016, 0.1584, 0.3830], + device='cuda:0'), in_proj_covar=tensor([0.0872, 0.0846, 0.0711, 0.0819, 0.0921, 0.0780, 0.0694, 0.0755], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 09:31:40,383 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.5192, 1.3034, 4.6973, 1.8136, 4.1096, 3.9308, 4.2122, 4.0613], + device='cuda:0'), covar=tensor([0.0483, 0.4442, 0.0391, 0.3253, 0.0970, 0.0770, 0.0497, 0.0613], + device='cuda:0'), in_proj_covar=tensor([0.0457, 0.0554, 0.0554, 0.0508, 0.0580, 0.0490, 0.0486, 0.0552], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 09:31:42,390 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80122.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:31:44,301 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=80124.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:31:50,381 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80133.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 09:31:50,991 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80134.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:31:54,314 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4206, 2.6571, 1.7905, 2.2773, 2.1215, 1.4039, 1.9936, 2.3125], + device='cuda:0'), covar=tensor([0.1529, 0.0384, 0.1218, 0.0679, 0.0687, 0.1646, 0.1109, 0.0813], + device='cuda:0'), in_proj_covar=tensor([0.0347, 0.0233, 0.0311, 0.0297, 0.0309, 0.0323, 0.0340, 0.0309], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 09:31:56,732 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-06 09:31:59,431 INFO [optim.py:369] (0/4) 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 09:32:00,147 INFO [train.py:901] (0/4) Epoch 10, batch 7400, loss[loss=0.2154, simple_loss=0.2832, pruned_loss=0.07384, over 7538.00 frames. ], tot_loss[loss=0.2403, simple_loss=0.3143, pruned_loss=0.08312, over 1611352.92 frames. ], batch size: 18, lr: 7.51e-03, grad_scale: 8.0 +2023-02-06 09:32:16,225 WARNING [train.py:1067] (0/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-06 09:32:34,304 INFO [train.py:901] (0/4) Epoch 10, batch 7450, loss[loss=0.2467, simple_loss=0.3326, pruned_loss=0.08047, over 8187.00 frames. ], tot_loss[loss=0.2408, simple_loss=0.3144, pruned_loss=0.08358, over 1607193.87 frames. ], batch size: 23, lr: 7.51e-03, grad_scale: 8.0 +2023-02-06 09:32:35,893 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.1828, 1.2133, 3.2916, 0.9739, 2.8449, 2.7272, 3.0083, 2.8861], + device='cuda:0'), covar=tensor([0.0741, 0.4070, 0.0766, 0.3611, 0.1390, 0.0984, 0.0762, 0.0884], + device='cuda:0'), in_proj_covar=tensor([0.0462, 0.0558, 0.0561, 0.0511, 0.0584, 0.0492, 0.0489, 0.0555], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 09:32:39,979 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6215, 1.9342, 3.3031, 1.3390, 2.4020, 2.0561, 1.6379, 2.1174], + device='cuda:0'), covar=tensor([0.1597, 0.2185, 0.0641, 0.3711, 0.1501, 0.2643, 0.1716, 0.2270], + device='cuda:0'), in_proj_covar=tensor([0.0488, 0.0511, 0.0532, 0.0574, 0.0614, 0.0554, 0.0467, 0.0606], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 09:32:54,360 WARNING [train.py:1067] (0/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-06 09:33:02,477 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80239.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:33:09,145 INFO [optim.py:369] (0/4) 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 09:33:09,863 INFO [train.py:901] (0/4) Epoch 10, batch 7500, loss[loss=0.2556, simple_loss=0.3169, pruned_loss=0.09714, over 7785.00 frames. ], tot_loss[loss=0.2421, simple_loss=0.3155, pruned_loss=0.08434, over 1610274.16 frames. ], batch size: 19, lr: 7.51e-03, grad_scale: 8.0 +2023-02-06 09:33:19,685 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 +2023-02-06 09:33:43,947 INFO [train.py:901] (0/4) Epoch 10, batch 7550, loss[loss=0.2224, simple_loss=0.3071, pruned_loss=0.06885, over 8035.00 frames. ], tot_loss[loss=0.2415, simple_loss=0.3152, pruned_loss=0.08386, over 1612156.94 frames. ], batch size: 22, lr: 7.50e-03, grad_scale: 8.0 +2023-02-06 09:33:46,241 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80318.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:34:15,027 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80343.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:34:16,558 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 +2023-02-06 09:34:17,459 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 10, batch 7600, loss[loss=0.2135, simple_loss=0.288, pruned_loss=0.06949, over 7543.00 frames. ], tot_loss[loss=0.2405, simple_loss=0.315, pruned_loss=0.083, over 1618060.42 frames. ], batch size: 18, lr: 7.50e-03, grad_scale: 8.0 +2023-02-06 09:34:22,311 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1100, 2.3682, 2.6460, 1.3531, 2.8024, 1.6944, 1.5384, 1.9717], + device='cuda:0'), covar=tensor([0.0468, 0.0243, 0.0206, 0.0476, 0.0240, 0.0537, 0.0559, 0.0298], + device='cuda:0'), in_proj_covar=tensor([0.0370, 0.0304, 0.0259, 0.0374, 0.0293, 0.0452, 0.0345, 0.0336], + device='cuda:0'), out_proj_covar=tensor([1.0690e-04, 8.5945e-05, 7.3445e-05, 1.0645e-04, 8.4249e-05, 1.3921e-04, + 9.9935e-05, 9.6759e-05], device='cuda:0') +2023-02-06 09:34:49,205 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 10, batch 7650, loss[loss=0.2842, simple_loss=0.3557, pruned_loss=0.1063, over 8240.00 frames. ], tot_loss[loss=0.2401, simple_loss=0.3145, pruned_loss=0.08285, over 1610111.13 frames. ], batch size: 24, lr: 7.50e-03, grad_scale: 8.0 +2023-02-06 09:35:03,327 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80411.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:35:06,125 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80415.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:35:20,856 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1680, 1.4994, 1.6634, 1.3662, 1.0467, 1.5093, 1.8724, 1.7370], + device='cuda:0'), covar=tensor([0.0494, 0.1329, 0.1736, 0.1419, 0.0619, 0.1514, 0.0683, 0.0572], + device='cuda:0'), in_proj_covar=tensor([0.0103, 0.0154, 0.0196, 0.0160, 0.0106, 0.0165, 0.0119, 0.0139], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0006], + device='cuda:0') +2023-02-06 09:35:27,279 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 10, batch 7700, loss[loss=0.2136, simple_loss=0.2911, pruned_loss=0.06806, over 8024.00 frames. ], tot_loss[loss=0.2394, simple_loss=0.3139, pruned_loss=0.08243, over 1610071.81 frames. ], batch size: 22, lr: 7.50e-03, grad_scale: 8.0 +2023-02-06 09:35:51,546 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80481.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:36:01,547 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80495.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:36:03,351 INFO [train.py:901] (0/4) Epoch 10, batch 7750, loss[loss=0.2049, simple_loss=0.2839, pruned_loss=0.06299, over 8235.00 frames. ], tot_loss[loss=0.2396, simple_loss=0.3135, pruned_loss=0.08283, over 1608146.48 frames. ], batch size: 22, lr: 7.49e-03, grad_scale: 8.0 +2023-02-06 09:36:06,760 WARNING [train.py:1067] (0/4) 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] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80512.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:36:15,240 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.1521, 1.2671, 4.3060, 1.5517, 3.8267, 3.5890, 3.8514, 3.7463], + device='cuda:0'), covar=tensor([0.0521, 0.4024, 0.0393, 0.3119, 0.0952, 0.0814, 0.0513, 0.0630], + device='cuda:0'), in_proj_covar=tensor([0.0471, 0.0566, 0.0564, 0.0519, 0.0589, 0.0498, 0.0496, 0.0564], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 09:36:18,625 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80520.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:36:19,249 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.2513, 1.3038, 3.3350, 0.9599, 2.9866, 2.8505, 3.0634, 2.9783], + device='cuda:0'), covar=tensor([0.0628, 0.3391, 0.0639, 0.3270, 0.1158, 0.0919, 0.0624, 0.0745], + device='cuda:0'), in_proj_covar=tensor([0.0470, 0.0564, 0.0561, 0.0517, 0.0587, 0.0496, 0.0494, 0.0562], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 09:36:23,373 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5870, 1.8730, 2.1705, 1.4887, 2.2654, 1.4251, 0.7331, 1.7779], + device='cuda:0'), covar=tensor([0.0432, 0.0233, 0.0159, 0.0314, 0.0228, 0.0609, 0.0547, 0.0211], + device='cuda:0'), in_proj_covar=tensor([0.0373, 0.0303, 0.0260, 0.0373, 0.0294, 0.0452, 0.0344, 0.0336], + device='cuda:0'), out_proj_covar=tensor([1.0778e-04, 8.5584e-05, 7.3774e-05, 1.0625e-04, 8.4444e-05, 1.3911e-04, + 9.9402e-05, 9.6612e-05], device='cuda:0') +2023-02-06 09:36:36,355 INFO [optim.py:369] (0/4) 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 09:36:37,066 INFO [train.py:901] (0/4) Epoch 10, batch 7800, loss[loss=0.2094, simple_loss=0.2916, pruned_loss=0.06358, over 8112.00 frames. ], tot_loss[loss=0.2396, simple_loss=0.3135, pruned_loss=0.08287, over 1612742.53 frames. ], batch size: 23, lr: 7.49e-03, grad_scale: 8.0 +2023-02-06 09:36:48,397 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.2597, 3.1721, 2.8682, 1.4095, 2.8614, 2.8830, 2.8999, 2.7106], + device='cuda:0'), covar=tensor([0.1015, 0.0752, 0.1308, 0.4923, 0.1123, 0.1160, 0.1522, 0.1050], + device='cuda:0'), in_proj_covar=tensor([0.0460, 0.0358, 0.0378, 0.0477, 0.0377, 0.0363, 0.0367, 0.0318], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 09:37:09,868 INFO [train.py:901] (0/4) Epoch 10, batch 7850, loss[loss=0.2438, simple_loss=0.3225, pruned_loss=0.08258, over 8473.00 frames. ], tot_loss[loss=0.2399, simple_loss=0.3137, pruned_loss=0.08303, over 1615871.31 frames. ], batch size: 25, lr: 7.49e-03, grad_scale: 8.0 +2023-02-06 09:37:12,391 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-06 09:37:40,857 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=80645.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:37:42,735 INFO [optim.py:369] (0/4) 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 09:37:43,434 INFO [train.py:901] (0/4) Epoch 10, batch 7900, loss[loss=0.2474, simple_loss=0.3319, pruned_loss=0.08143, over 8257.00 frames. ], tot_loss[loss=0.2408, simple_loss=0.3143, pruned_loss=0.08367, over 1614238.02 frames. ], batch size: 24, lr: 7.49e-03, grad_scale: 8.0 +2023-02-06 09:38:07,142 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.9933, 1.6298, 1.8181, 1.4885, 1.2799, 1.8257, 2.3016, 1.8651], + device='cuda:0'), covar=tensor([0.0439, 0.1211, 0.1697, 0.1425, 0.0566, 0.1406, 0.0605, 0.0626], + device='cuda:0'), in_proj_covar=tensor([0.0102, 0.0154, 0.0195, 0.0159, 0.0106, 0.0165, 0.0119, 0.0139], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0006], + device='cuda:0') +2023-02-06 09:38:16,940 INFO [train.py:901] (0/4) Epoch 10, batch 7950, loss[loss=0.2503, simple_loss=0.3, pruned_loss=0.1003, over 7192.00 frames. ], tot_loss[loss=0.24, simple_loss=0.3135, pruned_loss=0.08323, over 1615971.37 frames. ], batch size: 16, lr: 7.49e-03, grad_scale: 8.0 +2023-02-06 09:38:50,730 INFO [optim.py:369] (0/4) 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 09:38:51,443 INFO [train.py:901] (0/4) Epoch 10, batch 8000, loss[loss=0.2366, simple_loss=0.3227, pruned_loss=0.07521, over 8275.00 frames. ], tot_loss[loss=0.2382, simple_loss=0.3123, pruned_loss=0.08206, over 1615860.14 frames. ], batch size: 23, lr: 7.48e-03, grad_scale: 8.0 +2023-02-06 09:38:55,699 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80754.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:38:56,301 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=80755.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:38:59,809 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80760.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:39:20,341 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80791.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:39:24,998 INFO [train.py:901] (0/4) Epoch 10, batch 8050, loss[loss=0.3267, simple_loss=0.3741, pruned_loss=0.1397, over 7120.00 frames. ], tot_loss[loss=0.2378, simple_loss=0.3119, pruned_loss=0.08185, over 1606665.23 frames. ], batch size: 71, lr: 7.48e-03, grad_scale: 8.0 +2023-02-06 09:39:43,556 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=80825.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:39:45,321 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 +2023-02-06 09:39:48,194 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/epoch-10.pt +2023-02-06 09:39:58,264 WARNING [train.py:1067] (0/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-06 09:40:01,801 INFO [train.py:901] (0/4) Epoch 11, batch 0, loss[loss=0.2901, simple_loss=0.3512, pruned_loss=0.1145, over 8127.00 frames. ], tot_loss[loss=0.2901, simple_loss=0.3512, pruned_loss=0.1145, over 8127.00 frames. ], batch size: 22, lr: 7.14e-03, grad_scale: 8.0 +2023-02-06 09:40:01,802 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 09:40:13,094 INFO [train.py:935] (0/4) Epoch 11, validation: loss=0.1907, simple_loss=0.2907, pruned_loss=0.04534, over 944034.00 frames. +2023-02-06 09:40:13,095 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-06 09:40:23,924 INFO [optim.py:369] (0/4) 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 09:40:27,464 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-06 09:40:30,160 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=80856.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:40:39,882 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80870.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:40:47,958 INFO [train.py:901] (0/4) Epoch 11, batch 50, loss[loss=0.2047, simple_loss=0.2797, pruned_loss=0.06479, over 7807.00 frames. ], tot_loss[loss=0.2483, simple_loss=0.3233, pruned_loss=0.08664, over 368246.31 frames. ], batch size: 19, lr: 7.14e-03, grad_scale: 8.0 +2023-02-06 09:41:03,896 WARNING [train.py:1067] (0/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-06 09:41:24,358 INFO [train.py:901] (0/4) Epoch 11, batch 100, loss[loss=0.2022, simple_loss=0.2723, pruned_loss=0.06603, over 7522.00 frames. ], tot_loss[loss=0.2448, simple_loss=0.3203, pruned_loss=0.08466, over 646462.23 frames. ], batch size: 18, lr: 7.14e-03, grad_scale: 8.0 +2023-02-06 09:41:29,247 WARNING [train.py:1067] (0/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-06 09:41:30,742 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80940.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:41:32,066 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8825, 1.4988, 1.6183, 1.4112, 1.0092, 1.4375, 1.6067, 1.5338], + device='cuda:0'), covar=tensor([0.0472, 0.1172, 0.1653, 0.1325, 0.0574, 0.1439, 0.0677, 0.0577], + device='cuda:0'), in_proj_covar=tensor([0.0102, 0.0155, 0.0197, 0.0161, 0.0107, 0.0165, 0.0119, 0.0139], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0006], + device='cuda:0') +2023-02-06 09:41:32,119 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5105, 1.8488, 1.9515, 0.9502, 2.0541, 1.4659, 0.4280, 1.7198], + device='cuda:0'), covar=tensor([0.0328, 0.0207, 0.0154, 0.0319, 0.0193, 0.0529, 0.0543, 0.0147], + device='cuda:0'), in_proj_covar=tensor([0.0381, 0.0310, 0.0262, 0.0377, 0.0297, 0.0459, 0.0351, 0.0343], + device='cuda:0'), out_proj_covar=tensor([1.0991e-04, 8.7396e-05, 7.3964e-05, 1.0731e-04, 8.5085e-05, 1.4165e-04, + 1.0120e-04, 9.8394e-05], device='cuda:0') +2023-02-06 09:41:35,316 INFO [optim.py:369] (0/4) 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 09:41:51,862 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80971.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:41:58,405 INFO [train.py:901] (0/4) Epoch 11, batch 150, loss[loss=0.3116, simple_loss=0.3665, pruned_loss=0.1284, over 6803.00 frames. ], tot_loss[loss=0.2436, simple_loss=0.3187, pruned_loss=0.08424, over 864003.76 frames. ], batch size: 71, lr: 7.13e-03, grad_scale: 8.0 +2023-02-06 09:42:09,088 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2856, 2.2623, 1.7387, 2.0332, 1.8583, 1.4449, 1.8098, 1.7977], + device='cuda:0'), covar=tensor([0.1118, 0.0335, 0.0844, 0.0485, 0.0551, 0.1213, 0.0791, 0.0764], + device='cuda:0'), in_proj_covar=tensor([0.0345, 0.0234, 0.0312, 0.0298, 0.0305, 0.0326, 0.0338, 0.0305], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 09:42:14,053 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0089, 1.1022, 0.9948, 1.3047, 0.6578, 0.8849, 1.0475, 1.1419], + device='cuda:0'), covar=tensor([0.0690, 0.0654, 0.0860, 0.0570, 0.0914, 0.1148, 0.0576, 0.0583], + device='cuda:0'), in_proj_covar=tensor([0.0240, 0.0216, 0.0258, 0.0219, 0.0221, 0.0258, 0.0258, 0.0228], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 09:42:23,751 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81016.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:42:25,279 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 +2023-02-06 09:42:34,650 INFO [train.py:901] (0/4) Epoch 11, batch 200, loss[loss=0.2065, simple_loss=0.3012, pruned_loss=0.05589, over 8253.00 frames. ], tot_loss[loss=0.2434, simple_loss=0.318, pruned_loss=0.08443, over 1028823.00 frames. ], batch size: 24, lr: 7.13e-03, grad_scale: 8.0 +2023-02-06 09:42:37,132 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 +2023-02-06 09:42:43,008 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81041.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:42:47,015 INFO [optim.py:369] (0/4) 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 09:43:10,560 INFO [train.py:901] (0/4) Epoch 11, batch 250, loss[loss=0.2676, simple_loss=0.3187, pruned_loss=0.1083, over 5932.00 frames. ], tot_loss[loss=0.2406, simple_loss=0.3152, pruned_loss=0.08302, over 1158832.68 frames. ], batch size: 13, lr: 7.13e-03, grad_scale: 8.0 +2023-02-06 09:43:21,554 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-06 09:43:22,302 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=81098.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:43:31,363 WARNING [train.py:1067] (0/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-06 09:43:42,733 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81126.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:43:46,016 INFO [train.py:901] (0/4) Epoch 11, batch 300, loss[loss=0.2915, simple_loss=0.3541, pruned_loss=0.1145, over 8127.00 frames. ], tot_loss[loss=0.2391, simple_loss=0.3141, pruned_loss=0.08203, over 1263113.96 frames. ], batch size: 22, lr: 7.13e-03, grad_scale: 16.0 +2023-02-06 09:43:48,269 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81134.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:43:48,880 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=81135.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:43:57,132 INFO [optim.py:369] (0/4) 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 09:44:00,791 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81151.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:44:22,521 INFO [train.py:901] (0/4) Epoch 11, batch 350, loss[loss=0.2364, simple_loss=0.3028, pruned_loss=0.08498, over 8242.00 frames. ], tot_loss[loss=0.2392, simple_loss=0.3141, pruned_loss=0.08219, over 1343443.33 frames. ], batch size: 22, lr: 7.13e-03, grad_scale: 8.0 +2023-02-06 09:44:23,375 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9479, 3.8701, 2.5746, 2.6783, 2.7848, 2.1023, 2.8914, 3.1196], + device='cuda:0'), covar=tensor([0.1580, 0.0430, 0.0939, 0.0739, 0.0766, 0.1337, 0.1100, 0.0949], + device='cuda:0'), in_proj_covar=tensor([0.0344, 0.0232, 0.0308, 0.0297, 0.0303, 0.0324, 0.0337, 0.0304], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 09:44:33,038 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81196.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:44:44,351 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81213.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:44:45,750 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3125, 1.8919, 2.8033, 2.2126, 2.4566, 2.1510, 1.7490, 1.1710], + device='cuda:0'), covar=tensor([0.3888, 0.4011, 0.1080, 0.2477, 0.1833, 0.2131, 0.1668, 0.4064], + device='cuda:0'), in_proj_covar=tensor([0.0889, 0.0861, 0.0718, 0.0831, 0.0932, 0.0791, 0.0702, 0.0760], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 09:44:49,696 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81221.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:44:53,725 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81227.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:44:56,286 INFO [train.py:901] (0/4) Epoch 11, batch 400, loss[loss=0.1903, simple_loss=0.2796, pruned_loss=0.05054, over 8538.00 frames. ], tot_loss[loss=0.2409, simple_loss=0.3155, pruned_loss=0.08314, over 1404719.32 frames. ], batch size: 39, lr: 7.12e-03, grad_scale: 8.0 +2023-02-06 09:45:08,650 INFO [optim.py:369] (0/4) 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 09:45:10,270 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81250.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:45:11,720 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81252.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:45:32,911 INFO [train.py:901] (0/4) Epoch 11, batch 450, loss[loss=0.2649, simple_loss=0.3372, pruned_loss=0.09629, over 8515.00 frames. ], tot_loss[loss=0.2399, simple_loss=0.3139, pruned_loss=0.08294, over 1449371.17 frames. ], batch size: 28, lr: 7.12e-03, grad_scale: 8.0 +2023-02-06 09:45:57,096 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81317.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:45:58,453 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81319.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:46:06,251 INFO [train.py:901] (0/4) Epoch 11, batch 500, loss[loss=0.2552, simple_loss=0.3303, pruned_loss=0.09005, over 8460.00 frames. ], tot_loss[loss=0.2389, simple_loss=0.3131, pruned_loss=0.08235, over 1487075.84 frames. ], batch size: 29, lr: 7.12e-03, grad_scale: 8.0 +2023-02-06 09:46:17,545 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 11, batch 550, loss[loss=0.2163, simple_loss=0.2986, pruned_loss=0.06695, over 8286.00 frames. ], tot_loss[loss=0.2407, simple_loss=0.3144, pruned_loss=0.08351, over 1517090.13 frames. ], batch size: 23, lr: 7.12e-03, grad_scale: 8.0 +2023-02-06 09:46:59,741 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0077, 2.4118, 1.8773, 3.0135, 1.3203, 1.6584, 1.9749, 2.4456], + device='cuda:0'), covar=tensor([0.0764, 0.0890, 0.1011, 0.0344, 0.1176, 0.1477, 0.1050, 0.0726], + device='cuda:0'), in_proj_covar=tensor([0.0238, 0.0213, 0.0253, 0.0216, 0.0216, 0.0254, 0.0257, 0.0224], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 09:47:15,688 INFO [train.py:901] (0/4) Epoch 11, batch 600, loss[loss=0.2354, simple_loss=0.3243, pruned_loss=0.07324, over 8509.00 frames. ], tot_loss[loss=0.2378, simple_loss=0.3119, pruned_loss=0.08184, over 1536588.67 frames. ], batch size: 26, lr: 7.11e-03, grad_scale: 8.0 +2023-02-06 09:47:27,343 INFO [optim.py:369] (0/4) 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 09:47:35,461 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-06 09:47:42,482 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81469.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:47:48,462 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=81478.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:47:50,436 INFO [train.py:901] (0/4) Epoch 11, batch 650, loss[loss=0.2374, simple_loss=0.3025, pruned_loss=0.08608, over 7966.00 frames. ], tot_loss[loss=0.2387, simple_loss=0.3131, pruned_loss=0.08212, over 1558864.81 frames. ], batch size: 21, lr: 7.11e-03, grad_scale: 8.0 +2023-02-06 09:47:59,441 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81494.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:48:08,364 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81506.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:48:26,942 INFO [train.py:901] (0/4) Epoch 11, batch 700, loss[loss=0.2186, simple_loss=0.2843, pruned_loss=0.07648, over 7423.00 frames. ], tot_loss[loss=0.2376, simple_loss=0.3121, pruned_loss=0.08156, over 1568063.29 frames. ], batch size: 17, lr: 7.11e-03, grad_scale: 8.0 +2023-02-06 09:48:27,104 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81531.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:48:38,764 INFO [optim.py:369] (0/4) 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 09:48:48,477 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.2052, 1.1580, 3.3407, 0.9568, 2.9108, 2.7980, 3.0466, 2.9153], + device='cuda:0'), covar=tensor([0.0755, 0.3866, 0.0784, 0.3477, 0.1573, 0.1067, 0.0766, 0.0895], + device='cuda:0'), in_proj_covar=tensor([0.0475, 0.0572, 0.0577, 0.0526, 0.0604, 0.0509, 0.0502, 0.0576], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 09:49:01,395 INFO [train.py:901] (0/4) Epoch 11, batch 750, loss[loss=0.3008, simple_loss=0.3671, pruned_loss=0.1173, over 8767.00 frames. ], tot_loss[loss=0.2407, simple_loss=0.3148, pruned_loss=0.08328, over 1583142.60 frames. ], batch size: 30, lr: 7.11e-03, grad_scale: 8.0 +2023-02-06 09:49:07,228 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.53 vs. limit=5.0 +2023-02-06 09:49:10,215 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81593.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:49:24,851 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-06 09:49:33,713 WARNING [train.py:1067] (0/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-06 09:49:36,445 INFO [train.py:901] (0/4) Epoch 11, batch 800, loss[loss=0.2436, simple_loss=0.3286, pruned_loss=0.07934, over 8472.00 frames. ], tot_loss[loss=0.2381, simple_loss=0.3127, pruned_loss=0.08179, over 1590684.22 frames. ], batch size: 29, lr: 7.11e-03, grad_scale: 8.0 +2023-02-06 09:49:49,260 INFO [optim.py:369] (0/4) 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 09:49:58,255 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=81661.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:49:59,634 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=81663.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:50:11,544 INFO [train.py:901] (0/4) Epoch 11, batch 850, loss[loss=0.2382, simple_loss=0.3103, pruned_loss=0.083, over 7801.00 frames. ], tot_loss[loss=0.2381, simple_loss=0.3125, pruned_loss=0.08184, over 1595423.93 frames. ], batch size: 19, lr: 7.10e-03, grad_scale: 8.0 +2023-02-06 09:50:16,510 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81688.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:50:37,526 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9373, 1.6238, 2.2965, 1.7880, 2.0048, 1.8729, 1.4929, 0.6547], + device='cuda:0'), covar=tensor([0.3777, 0.3520, 0.1104, 0.2226, 0.1639, 0.2181, 0.1776, 0.3581], + device='cuda:0'), in_proj_covar=tensor([0.0877, 0.0849, 0.0714, 0.0825, 0.0920, 0.0785, 0.0698, 0.0754], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 09:50:46,071 INFO [train.py:901] (0/4) Epoch 11, batch 900, loss[loss=0.2192, simple_loss=0.2989, pruned_loss=0.06975, over 7150.00 frames. ], tot_loss[loss=0.2377, simple_loss=0.3123, pruned_loss=0.08157, over 1596655.94 frames. ], batch size: 71, lr: 7.10e-03, grad_scale: 8.0 +2023-02-06 09:50:58,849 INFO [optim.py:369] (0/4) 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 09:51:18,647 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81776.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:51:20,048 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81778.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:51:21,850 INFO [train.py:901] (0/4) Epoch 11, batch 950, loss[loss=0.2275, simple_loss=0.2932, pruned_loss=0.08087, over 7432.00 frames. ], tot_loss[loss=0.2372, simple_loss=0.312, pruned_loss=0.08122, over 1600370.74 frames. ], batch size: 17, lr: 7.10e-03, grad_scale: 8.0 +2023-02-06 09:51:22,935 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 +2023-02-06 09:51:24,920 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8955, 1.5760, 2.1936, 1.7641, 2.0069, 1.8277, 1.5073, 0.7126], + device='cuda:0'), covar=tensor([0.4079, 0.3828, 0.1153, 0.2338, 0.1621, 0.2163, 0.1622, 0.3739], + device='cuda:0'), in_proj_covar=tensor([0.0887, 0.0856, 0.0721, 0.0833, 0.0925, 0.0793, 0.0705, 0.0760], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 09:51:51,860 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-06 09:51:56,015 INFO [train.py:901] (0/4) Epoch 11, batch 1000, loss[loss=0.2061, simple_loss=0.2839, pruned_loss=0.06411, over 7653.00 frames. ], tot_loss[loss=0.2378, simple_loss=0.3128, pruned_loss=0.08137, over 1605844.64 frames. ], batch size: 19, lr: 7.10e-03, grad_scale: 8.0 +2023-02-06 09:52:07,439 INFO [optim.py:369] (0/4) 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 09:52:08,383 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81849.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:52:27,199 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-06 09:52:27,401 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81874.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:52:31,964 INFO [train.py:901] (0/4) Epoch 11, batch 1050, loss[loss=0.3133, simple_loss=0.3568, pruned_loss=0.1349, over 6678.00 frames. ], tot_loss[loss=0.2361, simple_loss=0.3114, pruned_loss=0.08042, over 1609714.79 frames. ], batch size: 71, lr: 7.09e-03, grad_scale: 8.0 +2023-02-06 09:52:39,058 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-06 09:53:06,143 INFO [train.py:901] (0/4) Epoch 11, batch 1100, loss[loss=0.2669, simple_loss=0.3269, pruned_loss=0.1035, over 8192.00 frames. ], tot_loss[loss=0.2368, simple_loss=0.3119, pruned_loss=0.08083, over 1613969.38 frames. ], batch size: 23, lr: 7.09e-03, grad_scale: 8.0 +2023-02-06 09:53:18,511 INFO [optim.py:369] (0/4) 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 09:53:41,605 INFO [train.py:901] (0/4) Epoch 11, batch 1150, loss[loss=0.2167, simple_loss=0.2881, pruned_loss=0.07268, over 7646.00 frames. ], tot_loss[loss=0.2357, simple_loss=0.3109, pruned_loss=0.08024, over 1615644.75 frames. ], batch size: 19, lr: 7.09e-03, grad_scale: 8.0 +2023-02-06 09:53:51,768 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-06 09:53:55,964 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-82000.pt +2023-02-06 09:54:00,447 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82005.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:54:17,960 INFO [train.py:901] (0/4) Epoch 11, batch 1200, loss[loss=0.2178, simple_loss=0.3033, pruned_loss=0.06614, over 8337.00 frames. ], tot_loss[loss=0.236, simple_loss=0.3108, pruned_loss=0.08064, over 1614758.98 frames. ], batch size: 25, lr: 7.09e-03, grad_scale: 8.0 +2023-02-06 09:54:18,717 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=82032.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:54:18,842 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82032.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:54:20,232 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82034.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:54:29,502 INFO [optim.py:369] (0/4) 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 09:54:36,802 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82057.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:54:38,208 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82059.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:54:47,997 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82073.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:54:48,015 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3485, 1.3826, 2.2564, 1.2201, 2.0749, 2.4523, 2.5937, 2.0715], + device='cuda:0'), covar=tensor([0.1009, 0.1182, 0.0488, 0.1880, 0.0658, 0.0371, 0.0584, 0.0746], + device='cuda:0'), in_proj_covar=tensor([0.0263, 0.0298, 0.0256, 0.0288, 0.0274, 0.0233, 0.0333, 0.0290], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 09:54:53,469 INFO [train.py:901] (0/4) Epoch 11, batch 1250, loss[loss=0.2067, simple_loss=0.2684, pruned_loss=0.07245, over 7716.00 frames. ], tot_loss[loss=0.2355, simple_loss=0.3108, pruned_loss=0.08012, over 1613894.93 frames. ], batch size: 18, lr: 7.09e-03, grad_scale: 8.0 +2023-02-06 09:55:29,353 INFO [train.py:901] (0/4) Epoch 11, batch 1300, loss[loss=0.2273, simple_loss=0.3089, pruned_loss=0.07285, over 8197.00 frames. ], tot_loss[loss=0.2356, simple_loss=0.3111, pruned_loss=0.07999, over 1614838.10 frames. ], batch size: 23, lr: 7.08e-03, grad_scale: 8.0 +2023-02-06 09:55:40,430 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=82147.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:55:40,863 INFO [optim.py:369] (0/4) 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 09:56:03,693 INFO [train.py:901] (0/4) Epoch 11, batch 1350, loss[loss=0.2798, simple_loss=0.3573, pruned_loss=0.1011, over 8195.00 frames. ], tot_loss[loss=0.2371, simple_loss=0.3122, pruned_loss=0.081, over 1614478.37 frames. ], batch size: 23, lr: 7.08e-03, grad_scale: 8.0 +2023-02-06 09:56:04,909 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 +2023-02-06 09:56:38,846 INFO [train.py:901] (0/4) Epoch 11, batch 1400, loss[loss=0.2405, simple_loss=0.3061, pruned_loss=0.08743, over 7909.00 frames. ], tot_loss[loss=0.2366, simple_loss=0.3117, pruned_loss=0.08074, over 1615256.89 frames. ], batch size: 20, lr: 7.08e-03, grad_scale: 8.0 +2023-02-06 09:56:51,062 INFO [optim.py:369] (0/4) 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 09:57:13,600 INFO [train.py:901] (0/4) Epoch 11, batch 1450, loss[loss=0.2332, simple_loss=0.3066, pruned_loss=0.07992, over 7525.00 frames. ], tot_loss[loss=0.2372, simple_loss=0.3125, pruned_loss=0.08094, over 1618297.50 frames. ], batch size: 18, lr: 7.08e-03, grad_scale: 8.0 +2023-02-06 09:57:17,215 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.1760, 1.2562, 1.4431, 1.2245, 0.7990, 1.3232, 1.1951, 0.8945], + device='cuda:0'), covar=tensor([0.0581, 0.1272, 0.1791, 0.1410, 0.0617, 0.1579, 0.0755, 0.0740], + device='cuda:0'), in_proj_covar=tensor([0.0101, 0.0153, 0.0195, 0.0158, 0.0105, 0.0165, 0.0118, 0.0139], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0006], + device='cuda:0') +2023-02-06 09:57:27,725 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-06 09:57:48,796 INFO [train.py:901] (0/4) Epoch 11, batch 1500, loss[loss=0.2283, simple_loss=0.3041, pruned_loss=0.07623, over 8657.00 frames. ], tot_loss[loss=0.2371, simple_loss=0.3122, pruned_loss=0.08094, over 1623511.42 frames. ], batch size: 34, lr: 7.08e-03, grad_scale: 8.0 +2023-02-06 09:58:01,382 INFO [optim.py:369] (0/4) 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 09:58:02,225 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=82349.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:58:10,585 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.1328, 1.2314, 1.2084, 0.6699, 1.2382, 1.0014, 0.1030, 1.1489], + device='cuda:0'), covar=tensor([0.0237, 0.0210, 0.0190, 0.0293, 0.0208, 0.0573, 0.0465, 0.0201], + device='cuda:0'), in_proj_covar=tensor([0.0383, 0.0315, 0.0265, 0.0378, 0.0301, 0.0464, 0.0353, 0.0349], + device='cuda:0'), out_proj_covar=tensor([1.1029e-04, 8.8593e-05, 7.4803e-05, 1.0718e-04, 8.5846e-05, 1.4273e-04, + 1.0163e-04, 9.9819e-05], device='cuda:0') +2023-02-06 09:58:13,167 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82364.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:58:24,626 INFO [train.py:901] (0/4) Epoch 11, batch 1550, loss[loss=0.239, simple_loss=0.3185, pruned_loss=0.07972, over 8290.00 frames. ], tot_loss[loss=0.2357, simple_loss=0.3113, pruned_loss=0.08, over 1624717.91 frames. ], batch size: 23, lr: 7.07e-03, grad_scale: 8.0 +2023-02-06 09:58:39,991 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82403.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:58:50,112 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=82417.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:58:57,861 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82428.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:58:59,757 INFO [train.py:901] (0/4) Epoch 11, batch 1600, loss[loss=0.1741, simple_loss=0.2502, pruned_loss=0.049, over 7264.00 frames. ], tot_loss[loss=0.2351, simple_loss=0.3103, pruned_loss=0.07995, over 1619932.67 frames. ], batch size: 16, lr: 7.07e-03, grad_scale: 8.0 +2023-02-06 09:59:13,003 INFO [optim.py:369] (0/4) 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 09:59:24,144 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=82464.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 09:59:26,842 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.8420, 5.8263, 5.1546, 2.2266, 5.1966, 5.5753, 5.4814, 5.2886], + device='cuda:0'), covar=tensor([0.0473, 0.0429, 0.0813, 0.4562, 0.0577, 0.0841, 0.1068, 0.0533], + device='cuda:0'), in_proj_covar=tensor([0.0466, 0.0375, 0.0384, 0.0482, 0.0377, 0.0368, 0.0369, 0.0324], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 09:59:36,415 INFO [train.py:901] (0/4) Epoch 11, batch 1650, loss[loss=0.21, simple_loss=0.2798, pruned_loss=0.07008, over 7824.00 frames. ], tot_loss[loss=0.2371, simple_loss=0.3119, pruned_loss=0.08114, over 1621159.96 frames. ], batch size: 20, lr: 7.07e-03, grad_scale: 8.0 +2023-02-06 09:59:37,935 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8159, 2.3639, 2.6931, 1.0929, 2.7229, 1.6321, 1.4171, 1.6673], + device='cuda:0'), covar=tensor([0.0755, 0.0291, 0.0248, 0.0619, 0.0300, 0.0693, 0.0778, 0.0450], + device='cuda:0'), in_proj_covar=tensor([0.0384, 0.0318, 0.0266, 0.0377, 0.0304, 0.0466, 0.0354, 0.0350], + device='cuda:0'), out_proj_covar=tensor([1.1041e-04, 8.9431e-05, 7.5022e-05, 1.0647e-04, 8.6759e-05, 1.4345e-04, + 1.0202e-04, 1.0024e-04], device='cuda:0') +2023-02-06 09:59:55,484 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 +2023-02-06 09:59:57,331 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82511.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 10:00:11,624 INFO [train.py:901] (0/4) Epoch 11, batch 1700, loss[loss=0.2442, simple_loss=0.3143, pruned_loss=0.08704, over 7969.00 frames. ], tot_loss[loss=0.2364, simple_loss=0.3109, pruned_loss=0.08097, over 1617810.07 frames. ], batch size: 21, lr: 7.07e-03, grad_scale: 8.0 +2023-02-06 10:00:12,460 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=82532.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:00:23,208 INFO [optim.py:369] (0/4) 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 10:00:35,356 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-02-06 10:00:47,536 INFO [train.py:901] (0/4) Epoch 11, batch 1750, loss[loss=0.2988, simple_loss=0.3554, pruned_loss=0.1211, over 7519.00 frames. ], tot_loss[loss=0.2377, simple_loss=0.3119, pruned_loss=0.08175, over 1614569.54 frames. ], batch size: 73, lr: 7.06e-03, grad_scale: 8.0 +2023-02-06 10:00:52,560 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82587.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:01:23,327 INFO [train.py:901] (0/4) Epoch 11, batch 1800, loss[loss=0.2166, simple_loss=0.2857, pruned_loss=0.07378, over 7982.00 frames. ], tot_loss[loss=0.2384, simple_loss=0.3126, pruned_loss=0.08211, over 1617419.21 frames. ], batch size: 21, lr: 7.06e-03, grad_scale: 8.0 +2023-02-06 10:01:35,825 INFO [optim.py:369] (0/4) 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 10:01:38,054 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.7403, 1.5265, 1.7221, 1.5755, 1.1254, 1.6938, 2.2783, 1.9811], + device='cuda:0'), covar=tensor([0.0481, 0.1334, 0.1787, 0.1468, 0.0651, 0.1551, 0.0674, 0.0603], + device='cuda:0'), in_proj_covar=tensor([0.0102, 0.0155, 0.0198, 0.0160, 0.0106, 0.0166, 0.0119, 0.0140], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0006], + device='cuda:0') +2023-02-06 10:01:46,763 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5942, 1.4966, 1.5998, 1.5445, 0.9438, 1.5248, 1.5397, 1.3168], + device='cuda:0'), covar=tensor([0.0494, 0.1081, 0.1617, 0.1249, 0.0553, 0.1311, 0.0645, 0.0632], + device='cuda:0'), in_proj_covar=tensor([0.0102, 0.0154, 0.0197, 0.0160, 0.0106, 0.0165, 0.0119, 0.0140], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0006], + device='cuda:0') +2023-02-06 10:01:58,587 INFO [train.py:901] (0/4) Epoch 11, batch 1850, loss[loss=0.2351, simple_loss=0.3148, pruned_loss=0.07771, over 8478.00 frames. ], tot_loss[loss=0.2375, simple_loss=0.3119, pruned_loss=0.08158, over 1617211.83 frames. ], batch size: 27, lr: 7.06e-03, grad_scale: 8.0 +2023-02-06 10:02:03,664 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7711, 2.3617, 1.8110, 2.2017, 2.0816, 1.6667, 2.0122, 2.1947], + device='cuda:0'), covar=tensor([0.1001, 0.0269, 0.0822, 0.0446, 0.0509, 0.1021, 0.0651, 0.0628], + device='cuda:0'), in_proj_covar=tensor([0.0348, 0.0231, 0.0311, 0.0297, 0.0300, 0.0322, 0.0339, 0.0306], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 10:02:18,370 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=82708.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:02:26,580 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82720.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:02:34,008 INFO [train.py:901] (0/4) Epoch 11, batch 1900, loss[loss=0.2522, simple_loss=0.3281, pruned_loss=0.08818, over 8292.00 frames. ], tot_loss[loss=0.237, simple_loss=0.3116, pruned_loss=0.08115, over 1614689.26 frames. ], batch size: 23, lr: 7.06e-03, grad_scale: 8.0 +2023-02-06 10:02:43,683 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82745.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:02:45,529 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 11, batch 1950, loss[loss=0.2233, simple_loss=0.311, pruned_loss=0.06778, over 8453.00 frames. ], tot_loss[loss=0.2367, simple_loss=0.3112, pruned_loss=0.08114, over 1610964.71 frames. ], batch size: 25, lr: 7.06e-03, grad_scale: 8.0 +2023-02-06 10:03:12,343 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-06 10:03:13,856 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82788.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:03:17,446 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.69 vs. limit=5.0 +2023-02-06 10:03:26,482 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-06 10:03:32,226 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82813.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:03:39,063 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=82823.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:03:39,129 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6467, 1.8710, 2.9829, 1.3653, 2.2699, 1.9807, 1.6678, 2.1301], + device='cuda:0'), covar=tensor([0.1472, 0.2164, 0.0578, 0.3580, 0.1376, 0.2453, 0.1683, 0.1961], + device='cuda:0'), in_proj_covar=tensor([0.0481, 0.0512, 0.0534, 0.0575, 0.0614, 0.0546, 0.0469, 0.0609], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 10:03:45,036 INFO [train.py:901] (0/4) Epoch 11, batch 2000, loss[loss=0.2076, simple_loss=0.2747, pruned_loss=0.07022, over 7635.00 frames. ], tot_loss[loss=0.2364, simple_loss=0.311, pruned_loss=0.08092, over 1608719.47 frames. ], batch size: 19, lr: 7.05e-03, grad_scale: 8.0 +2023-02-06 10:03:47,025 WARNING [train.py:1067] (0/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-06 10:03:56,665 INFO [optim.py:369] (0/4) 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 10:04:01,551 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=82855.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 10:04:19,321 INFO [train.py:901] (0/4) Epoch 11, batch 2050, loss[loss=0.2218, simple_loss=0.3118, pruned_loss=0.06587, over 8334.00 frames. ], tot_loss[loss=0.2366, simple_loss=0.3115, pruned_loss=0.08086, over 1613148.71 frames. ], batch size: 26, lr: 7.05e-03, grad_scale: 8.0 +2023-02-06 10:04:27,382 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([6.0123, 6.1113, 5.1724, 2.5641, 5.3942, 5.8027, 5.5809, 5.4182], + device='cuda:0'), covar=tensor([0.0560, 0.0374, 0.0825, 0.4793, 0.0667, 0.0568, 0.0969, 0.0553], + device='cuda:0'), in_proj_covar=tensor([0.0465, 0.0375, 0.0383, 0.0483, 0.0375, 0.0366, 0.0370, 0.0324], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 10:04:55,307 INFO [train.py:901] (0/4) Epoch 11, batch 2100, loss[loss=0.2359, simple_loss=0.3062, pruned_loss=0.08281, over 8236.00 frames. ], tot_loss[loss=0.2359, simple_loss=0.3108, pruned_loss=0.08049, over 1612674.65 frames. ], batch size: 22, lr: 7.05e-03, grad_scale: 8.0 +2023-02-06 10:04:55,375 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 10:05:13,279 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0114, 1.4523, 1.6101, 1.3097, 0.8966, 1.5101, 1.7381, 1.6392], + device='cuda:0'), covar=tensor([0.0486, 0.1249, 0.1609, 0.1361, 0.0608, 0.1471, 0.0696, 0.0549], + device='cuda:0'), in_proj_covar=tensor([0.0102, 0.0154, 0.0195, 0.0160, 0.0105, 0.0166, 0.0118, 0.0139], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0006], + device='cuda:0') +2023-02-06 10:05:22,081 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=82970.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 10:05:29,198 INFO [train.py:901] (0/4) Epoch 11, batch 2150, loss[loss=0.2662, simple_loss=0.3403, pruned_loss=0.09606, over 8360.00 frames. ], tot_loss[loss=0.2368, simple_loss=0.3119, pruned_loss=0.08088, over 1613028.02 frames. ], batch size: 24, lr: 7.05e-03, grad_scale: 8.0 +2023-02-06 10:05:29,366 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82981.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:05:49,411 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-02-06 10:06:04,060 INFO [train.py:901] (0/4) Epoch 11, batch 2200, loss[loss=0.2374, simple_loss=0.3079, pruned_loss=0.08341, over 7803.00 frames. ], tot_loss[loss=0.2353, simple_loss=0.3102, pruned_loss=0.08022, over 1608784.25 frames. ], batch size: 20, lr: 7.05e-03, grad_scale: 8.0 +2023-02-06 10:06:15,765 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83046.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:06:16,954 INFO [optim.py:369] (0/4) 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] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83079.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:06:40,019 INFO [train.py:901] (0/4) Epoch 11, batch 2250, loss[loss=0.2464, simple_loss=0.3251, pruned_loss=0.08379, over 7919.00 frames. ], tot_loss[loss=0.2355, simple_loss=0.3101, pruned_loss=0.0804, over 1605172.93 frames. ], batch size: 20, lr: 7.04e-03, grad_scale: 8.0 +2023-02-06 10:06:46,975 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4926, 2.6059, 1.7956, 2.1851, 2.0849, 1.4930, 2.1856, 2.2330], + device='cuda:0'), covar=tensor([0.1518, 0.0408, 0.1167, 0.0724, 0.0753, 0.1460, 0.0927, 0.0983], + device='cuda:0'), in_proj_covar=tensor([0.0351, 0.0236, 0.0315, 0.0297, 0.0304, 0.0326, 0.0341, 0.0308], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 10:06:49,943 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 +2023-02-06 10:06:55,778 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 11, batch 2300, loss[loss=0.2254, simple_loss=0.286, pruned_loss=0.08243, over 7433.00 frames. ], tot_loss[loss=0.2367, simple_loss=0.3115, pruned_loss=0.08098, over 1611182.52 frames. ], batch size: 17, lr: 7.04e-03, grad_scale: 8.0 +2023-02-06 10:07:25,656 INFO [optim.py:369] (0/4) 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 10:07:48,922 INFO [train.py:901] (0/4) Epoch 11, batch 2350, loss[loss=0.2477, simple_loss=0.3228, pruned_loss=0.08626, over 8485.00 frames. ], tot_loss[loss=0.2373, simple_loss=0.3121, pruned_loss=0.08125, over 1615122.34 frames. ], batch size: 29, lr: 7.04e-03, grad_scale: 16.0 +2023-02-06 10:08:12,070 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=83214.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:08:20,033 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83226.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 10:08:23,038 INFO [train.py:901] (0/4) Epoch 11, batch 2400, loss[loss=0.2502, simple_loss=0.3182, pruned_loss=0.0911, over 8365.00 frames. ], tot_loss[loss=0.2371, simple_loss=0.3123, pruned_loss=0.08095, over 1618669.69 frames. ], batch size: 24, lr: 7.04e-03, grad_scale: 16.0 +2023-02-06 10:08:35,105 INFO [optim.py:369] (0/4) 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 10:08:37,251 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83251.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 10:08:58,645 INFO [train.py:901] (0/4) Epoch 11, batch 2450, loss[loss=0.2393, simple_loss=0.3083, pruned_loss=0.08513, over 7942.00 frames. ], tot_loss[loss=0.235, simple_loss=0.3107, pruned_loss=0.07961, over 1618729.32 frames. ], batch size: 20, lr: 7.04e-03, grad_scale: 16.0 +2023-02-06 10:09:06,343 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6355, 2.1577, 4.3528, 1.3459, 2.9724, 2.2010, 1.5621, 2.7127], + device='cuda:0'), covar=tensor([0.1742, 0.2366, 0.0781, 0.4070, 0.1663, 0.2803, 0.2015, 0.2450], + device='cuda:0'), in_proj_covar=tensor([0.0481, 0.0507, 0.0532, 0.0569, 0.0611, 0.0543, 0.0464, 0.0603], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 10:09:13,729 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=83302.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:09:13,795 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83302.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:09:29,007 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=83325.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:09:30,521 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83327.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:09:32,764 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-02-06 10:09:33,050 INFO [train.py:901] (0/4) Epoch 11, batch 2500, loss[loss=0.261, simple_loss=0.3332, pruned_loss=0.09436, over 8336.00 frames. ], tot_loss[loss=0.2351, simple_loss=0.3107, pruned_loss=0.07979, over 1623750.39 frames. ], batch size: 26, lr: 7.03e-03, grad_scale: 16.0 +2023-02-06 10:09:44,607 INFO [optim.py:369] (0/4) 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 10:10:07,382 INFO [train.py:901] (0/4) Epoch 11, batch 2550, loss[loss=0.1829, simple_loss=0.2647, pruned_loss=0.05056, over 7923.00 frames. ], tot_loss[loss=0.2357, simple_loss=0.311, pruned_loss=0.08024, over 1621693.66 frames. ], batch size: 20, lr: 7.03e-03, grad_scale: 16.0 +2023-02-06 10:10:28,997 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5281, 1.4701, 2.7830, 1.2972, 1.8806, 2.9557, 3.1017, 2.5180], + device='cuda:0'), covar=tensor([0.1107, 0.1456, 0.0390, 0.1974, 0.0964, 0.0315, 0.0453, 0.0678], + device='cuda:0'), in_proj_covar=tensor([0.0259, 0.0297, 0.0254, 0.0288, 0.0271, 0.0234, 0.0333, 0.0287], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-06 10:10:43,102 INFO [train.py:901] (0/4) Epoch 11, batch 2600, loss[loss=0.2461, simple_loss=0.3236, pruned_loss=0.08428, over 8348.00 frames. ], tot_loss[loss=0.2368, simple_loss=0.3115, pruned_loss=0.081, over 1615320.44 frames. ], batch size: 24, lr: 7.03e-03, grad_scale: 16.0 +2023-02-06 10:10:48,776 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1694, 1.4787, 4.3257, 1.9869, 2.4853, 5.0331, 4.9988, 4.3735], + device='cuda:0'), covar=tensor([0.1017, 0.1648, 0.0254, 0.1818, 0.0956, 0.0144, 0.0260, 0.0496], + device='cuda:0'), in_proj_covar=tensor([0.0258, 0.0295, 0.0253, 0.0286, 0.0268, 0.0232, 0.0331, 0.0285], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002], + device='cuda:0') +2023-02-06 10:10:49,493 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83440.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:10:54,710 INFO [optim.py:369] (0/4) 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 10:11:17,502 INFO [train.py:901] (0/4) Epoch 11, batch 2650, loss[loss=0.2124, simple_loss=0.2874, pruned_loss=0.0687, over 8546.00 frames. ], tot_loss[loss=0.2366, simple_loss=0.3112, pruned_loss=0.08104, over 1616756.18 frames. ], batch size: 31, lr: 7.03e-03, grad_scale: 16.0 +2023-02-06 10:11:52,414 INFO [train.py:901] (0/4) Epoch 11, batch 2700, loss[loss=0.2859, simple_loss=0.3399, pruned_loss=0.1159, over 8687.00 frames. ], tot_loss[loss=0.2382, simple_loss=0.3127, pruned_loss=0.08189, over 1620349.64 frames. ], batch size: 49, lr: 7.02e-03, grad_scale: 16.0 +2023-02-06 10:12:04,665 INFO [optim.py:369] (0/4) 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 10:12:11,619 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=83558.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:12:18,882 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 +2023-02-06 10:12:27,317 INFO [train.py:901] (0/4) Epoch 11, batch 2750, loss[loss=0.2336, simple_loss=0.3214, pruned_loss=0.07288, over 8259.00 frames. ], tot_loss[loss=0.2367, simple_loss=0.3115, pruned_loss=0.08095, over 1619557.13 frames. ], batch size: 24, lr: 7.02e-03, grad_scale: 16.0 +2023-02-06 10:13:03,328 INFO [train.py:901] (0/4) Epoch 11, batch 2800, loss[loss=0.2446, simple_loss=0.3213, pruned_loss=0.08399, over 8483.00 frames. ], tot_loss[loss=0.2385, simple_loss=0.3128, pruned_loss=0.08205, over 1620659.66 frames. ], batch size: 28, lr: 7.02e-03, grad_scale: 16.0 +2023-02-06 10:13:10,443 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8195, 1.4313, 1.6903, 1.3551, 0.9673, 1.4979, 1.5840, 1.4007], + device='cuda:0'), covar=tensor([0.0523, 0.1247, 0.1610, 0.1367, 0.0598, 0.1473, 0.0701, 0.0634], + device='cuda:0'), in_proj_covar=tensor([0.0101, 0.0154, 0.0194, 0.0159, 0.0105, 0.0163, 0.0117, 0.0139], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0006], + device='cuda:0') +2023-02-06 10:13:13,810 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=83646.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:13:15,057 INFO [optim.py:369] (0/4) 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 10:13:32,689 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83673.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:13:37,608 INFO [train.py:901] (0/4) Epoch 11, batch 2850, loss[loss=0.2639, simple_loss=0.3271, pruned_loss=0.1004, over 7235.00 frames. ], tot_loss[loss=0.2383, simple_loss=0.3126, pruned_loss=0.082, over 1621803.42 frames. ], batch size: 16, lr: 7.02e-03, grad_scale: 16.0 +2023-02-06 10:13:47,886 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83696.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:14:04,215 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1178, 2.4948, 1.8387, 3.0949, 1.4516, 1.5489, 1.9866, 2.4933], + device='cuda:0'), covar=tensor([0.0820, 0.0767, 0.1101, 0.0358, 0.1238, 0.1611, 0.1161, 0.0864], + device='cuda:0'), in_proj_covar=tensor([0.0241, 0.0217, 0.0256, 0.0218, 0.0217, 0.0254, 0.0259, 0.0226], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 10:14:05,191 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.59 vs. limit=5.0 +2023-02-06 10:14:05,621 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83721.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:14:12,967 INFO [train.py:901] (0/4) Epoch 11, batch 2900, loss[loss=0.1889, simple_loss=0.2587, pruned_loss=0.05952, over 7726.00 frames. ], tot_loss[loss=0.2383, simple_loss=0.3126, pruned_loss=0.08202, over 1620948.22 frames. ], batch size: 18, lr: 7.02e-03, grad_scale: 16.0 +2023-02-06 10:14:15,839 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7728, 1.4085, 1.6571, 1.3134, 1.0185, 1.4624, 1.6845, 1.4526], + device='cuda:0'), covar=tensor([0.0495, 0.1203, 0.1569, 0.1338, 0.0561, 0.1465, 0.0623, 0.0627], + device='cuda:0'), in_proj_covar=tensor([0.0101, 0.0153, 0.0194, 0.0158, 0.0105, 0.0163, 0.0116, 0.0138], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0006], + device='cuda:0') +2023-02-06 10:14:25,266 INFO [optim.py:369] (0/4) 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 10:14:34,258 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83761.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:14:46,310 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=83778.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:14:48,176 INFO [train.py:901] (0/4) Epoch 11, batch 2950, loss[loss=0.2471, simple_loss=0.3261, pruned_loss=0.08409, over 8476.00 frames. ], tot_loss[loss=0.2378, simple_loss=0.3126, pruned_loss=0.0815, over 1619770.27 frames. ], batch size: 29, lr: 7.01e-03, grad_scale: 16.0 +2023-02-06 10:14:53,615 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-06 10:15:09,068 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 +2023-02-06 10:15:18,759 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.90 vs. limit=5.0 +2023-02-06 10:15:22,306 INFO [train.py:901] (0/4) Epoch 11, batch 3000, loss[loss=0.2439, simple_loss=0.3127, pruned_loss=0.08751, over 8028.00 frames. ], tot_loss[loss=0.2385, simple_loss=0.3131, pruned_loss=0.08196, over 1619704.62 frames. ], batch size: 22, lr: 7.01e-03, grad_scale: 16.0 +2023-02-06 10:15:22,306 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 10:15:34,552 INFO [train.py:935] (0/4) Epoch 11, validation: loss=0.1889, simple_loss=0.2886, pruned_loss=0.04461, over 944034.00 frames. +2023-02-06 10:15:34,552 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-06 10:15:46,617 INFO [optim.py:369] (0/4) 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 10:16:10,350 INFO [train.py:901] (0/4) Epoch 11, batch 3050, loss[loss=0.2301, simple_loss=0.2997, pruned_loss=0.08022, over 8232.00 frames. ], tot_loss[loss=0.2388, simple_loss=0.3137, pruned_loss=0.08192, over 1618623.30 frames. ], batch size: 22, lr: 7.01e-03, grad_scale: 16.0 +2023-02-06 10:16:17,990 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.34 vs. limit=5.0 +2023-02-06 10:16:40,426 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3306, 1.9166, 2.9913, 2.3568, 2.5932, 2.1747, 1.7031, 1.3285], + device='cuda:0'), covar=tensor([0.3965, 0.4296, 0.1221, 0.2572, 0.2053, 0.2229, 0.1671, 0.4421], + device='cuda:0'), in_proj_covar=tensor([0.0887, 0.0865, 0.0727, 0.0837, 0.0938, 0.0796, 0.0703, 0.0766], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 10:16:43,140 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83929.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:16:44,274 INFO [train.py:901] (0/4) Epoch 11, batch 3100, loss[loss=0.2248, simple_loss=0.304, pruned_loss=0.07284, over 8247.00 frames. ], tot_loss[loss=0.2371, simple_loss=0.3123, pruned_loss=0.0809, over 1617477.03 frames. ], batch size: 22, lr: 7.01e-03, grad_scale: 16.0 +2023-02-06 10:16:45,446 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.84 vs. limit=5.0 +2023-02-06 10:16:55,423 INFO [optim.py:369] (0/4) 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 10:17:00,084 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83954.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:17:08,761 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=83967.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:17:18,436 INFO [train.py:901] (0/4) Epoch 11, batch 3150, loss[loss=0.2024, simple_loss=0.2631, pruned_loss=0.07089, over 7220.00 frames. ], tot_loss[loss=0.2371, simple_loss=0.3121, pruned_loss=0.08104, over 1613485.68 frames. ], batch size: 16, lr: 7.01e-03, grad_scale: 16.0 +2023-02-06 10:17:31,199 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-84000.pt +2023-02-06 10:17:34,193 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84003.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:17:44,259 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84017.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:17:53,249 INFO [train.py:901] (0/4) Epoch 11, batch 3200, loss[loss=0.2033, simple_loss=0.2729, pruned_loss=0.06685, over 7723.00 frames. ], tot_loss[loss=0.2376, simple_loss=0.3123, pruned_loss=0.08145, over 1615310.37 frames. ], batch size: 18, lr: 7.00e-03, grad_scale: 8.0 +2023-02-06 10:18:01,369 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84042.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:18:05,761 INFO [optim.py:369] (0/4) 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 10:18:27,179 INFO [train.py:901] (0/4) Epoch 11, batch 3250, loss[loss=0.245, simple_loss=0.3194, pruned_loss=0.08535, over 7442.00 frames. ], tot_loss[loss=0.2379, simple_loss=0.3126, pruned_loss=0.08167, over 1614302.29 frames. ], batch size: 17, lr: 7.00e-03, grad_scale: 8.0 +2023-02-06 10:18:50,412 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84115.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:18:53,044 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.6924, 4.5378, 4.1167, 2.0301, 4.0614, 4.2746, 4.3091, 3.8789], + device='cuda:0'), covar=tensor([0.0682, 0.0601, 0.1026, 0.4908, 0.0836, 0.0797, 0.1087, 0.0820], + device='cuda:0'), in_proj_covar=tensor([0.0465, 0.0375, 0.0380, 0.0482, 0.0377, 0.0372, 0.0375, 0.0330], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 10:18:55,063 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=84122.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:19:01,828 INFO [train.py:901] (0/4) Epoch 11, batch 3300, loss[loss=0.2058, simple_loss=0.2828, pruned_loss=0.06439, over 7817.00 frames. ], tot_loss[loss=0.2371, simple_loss=0.3119, pruned_loss=0.08119, over 1615779.69 frames. ], batch size: 20, lr: 7.00e-03, grad_scale: 8.0 +2023-02-06 10:19:13,369 INFO [optim.py:369] (0/4) 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 10:19:35,412 INFO [train.py:901] (0/4) Epoch 11, batch 3350, loss[loss=0.2503, simple_loss=0.3185, pruned_loss=0.09108, over 8238.00 frames. ], tot_loss[loss=0.2374, simple_loss=0.3116, pruned_loss=0.08163, over 1612848.75 frames. ], batch size: 22, lr: 7.00e-03, grad_scale: 8.0 +2023-02-06 10:19:41,345 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 +2023-02-06 10:20:01,008 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 11, batch 3400, loss[loss=0.2279, simple_loss=0.3085, pruned_loss=0.07372, over 8315.00 frames. ], tot_loss[loss=0.2362, simple_loss=0.3106, pruned_loss=0.08087, over 1610268.89 frames. ], batch size: 26, lr: 7.00e-03, grad_scale: 8.0 +2023-02-06 10:20:10,593 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-02-06 10:20:15,223 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84237.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:20:15,907 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8781, 1.3001, 1.4707, 1.2026, 0.9315, 1.3322, 1.6306, 1.4552], + device='cuda:0'), covar=tensor([0.0505, 0.1242, 0.1735, 0.1485, 0.0590, 0.1507, 0.0695, 0.0619], + device='cuda:0'), in_proj_covar=tensor([0.0100, 0.0153, 0.0190, 0.0157, 0.0104, 0.0163, 0.0116, 0.0138], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0006, 0.0006], + device='cuda:0') +2023-02-06 10:20:22,012 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8221, 2.0202, 2.0279, 1.6168, 2.1082, 1.7181, 1.1872, 1.8137], + device='cuda:0'), covar=tensor([0.0311, 0.0187, 0.0129, 0.0261, 0.0224, 0.0394, 0.0395, 0.0167], + device='cuda:0'), in_proj_covar=tensor([0.0389, 0.0323, 0.0267, 0.0386, 0.0309, 0.0471, 0.0358, 0.0352], + device='cuda:0'), out_proj_covar=tensor([1.1151e-04, 9.0463e-05, 7.4865e-05, 1.0915e-04, 8.7875e-05, 1.4464e-04, + 1.0276e-04, 1.0047e-04], device='cuda:0') +2023-02-06 10:20:23,163 INFO [optim.py:369] (0/4) 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 10:20:26,648 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84254.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:20:45,375 INFO [train.py:901] (0/4) Epoch 11, batch 3450, loss[loss=0.2274, simple_loss=0.3095, pruned_loss=0.07265, over 8360.00 frames. ], tot_loss[loss=0.2378, simple_loss=0.3123, pruned_loss=0.08161, over 1612176.11 frames. ], batch size: 24, lr: 6.99e-03, grad_scale: 8.0 +2023-02-06 10:21:06,413 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=84311.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:21:20,272 INFO [train.py:901] (0/4) Epoch 11, batch 3500, loss[loss=0.2689, simple_loss=0.334, pruned_loss=0.1019, over 8660.00 frames. ], tot_loss[loss=0.239, simple_loss=0.3135, pruned_loss=0.08219, over 1615231.40 frames. ], batch size: 34, lr: 6.99e-03, grad_scale: 8.0 +2023-02-06 10:21:31,064 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=84347.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:21:32,263 INFO [optim.py:369] (0/4) 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 10:21:36,444 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84354.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:21:41,538 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 +2023-02-06 10:21:48,760 WARNING [train.py:1067] (0/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-06 10:21:54,128 INFO [train.py:901] (0/4) Epoch 11, batch 3550, loss[loss=0.181, simple_loss=0.2608, pruned_loss=0.05063, over 7803.00 frames. ], tot_loss[loss=0.2369, simple_loss=0.3119, pruned_loss=0.08098, over 1614284.47 frames. ], batch size: 19, lr: 6.99e-03, grad_scale: 8.0 +2023-02-06 10:22:25,834 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84426.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:22:28,976 INFO [train.py:901] (0/4) Epoch 11, batch 3600, loss[loss=0.3101, simple_loss=0.3645, pruned_loss=0.1278, over 8326.00 frames. ], tot_loss[loss=0.2379, simple_loss=0.3125, pruned_loss=0.08162, over 1614245.14 frames. ], batch size: 26, lr: 6.99e-03, grad_scale: 8.0 +2023-02-06 10:22:41,783 INFO [optim.py:369] (0/4) 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 10:22:48,586 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=84459.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:22:50,733 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84462.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:23:01,848 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.8936, 2.6866, 3.4361, 2.1220, 1.7251, 3.4832, 0.4388, 2.0638], + device='cuda:0'), covar=tensor([0.2358, 0.1570, 0.0358, 0.2596, 0.4679, 0.0546, 0.4283, 0.2125], + device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0165, 0.0097, 0.0212, 0.0252, 0.0104, 0.0165, 0.0165], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 10:23:03,524 INFO [train.py:901] (0/4) Epoch 11, batch 3650, loss[loss=0.206, simple_loss=0.2724, pruned_loss=0.06983, over 7539.00 frames. ], tot_loss[loss=0.2376, simple_loss=0.3122, pruned_loss=0.08146, over 1613420.08 frames. ], batch size: 18, lr: 6.99e-03, grad_scale: 8.0 +2023-02-06 10:23:11,594 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84493.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:23:28,880 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84518.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:23:37,222 INFO [train.py:901] (0/4) Epoch 11, batch 3700, loss[loss=0.2199, simple_loss=0.3067, pruned_loss=0.06661, over 8189.00 frames. ], tot_loss[loss=0.2376, simple_loss=0.3123, pruned_loss=0.08141, over 1618018.94 frames. ], batch size: 23, lr: 6.98e-03, grad_scale: 8.0 +2023-02-06 10:23:46,128 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84543.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:23:48,580 WARNING [train.py:1067] (0/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-06 10:23:49,862 INFO [optim.py:369] (0/4) 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 10:23:56,591 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8158, 2.0653, 2.3160, 1.1376, 2.3625, 1.6202, 0.7484, 1.8873], + device='cuda:0'), covar=tensor([0.0401, 0.0258, 0.0182, 0.0417, 0.0248, 0.0570, 0.0554, 0.0213], + device='cuda:0'), in_proj_covar=tensor([0.0381, 0.0318, 0.0265, 0.0378, 0.0304, 0.0466, 0.0354, 0.0346], + device='cuda:0'), out_proj_covar=tensor([1.0910e-04, 8.9188e-05, 7.3993e-05, 1.0689e-04, 8.6500e-05, 1.4288e-04, + 1.0138e-04, 9.8619e-05], device='cuda:0') +2023-02-06 10:23:58,476 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=84561.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:24:07,283 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84574.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:24:11,838 INFO [train.py:901] (0/4) Epoch 11, batch 3750, loss[loss=0.2516, simple_loss=0.3259, pruned_loss=0.08861, over 8344.00 frames. ], tot_loss[loss=0.2373, simple_loss=0.3118, pruned_loss=0.08139, over 1614182.63 frames. ], batch size: 26, lr: 6.98e-03, grad_scale: 8.0 +2023-02-06 10:24:23,981 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 11, batch 3800, loss[loss=0.2317, simple_loss=0.3063, pruned_loss=0.07854, over 8111.00 frames. ], tot_loss[loss=0.2374, simple_loss=0.3115, pruned_loss=0.08165, over 1609319.03 frames. ], batch size: 23, lr: 6.98e-03, grad_scale: 8.0 +2023-02-06 10:24:58,773 INFO [optim.py:369] (0/4) 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 10:25:00,927 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84651.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:25:02,488 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.85 vs. limit=5.0 +2023-02-06 10:25:18,150 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84676.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:25:21,362 INFO [train.py:901] (0/4) Epoch 11, batch 3850, loss[loss=0.2399, simple_loss=0.3106, pruned_loss=0.08456, over 7799.00 frames. ], tot_loss[loss=0.2384, simple_loss=0.3124, pruned_loss=0.08218, over 1610235.32 frames. ], batch size: 20, lr: 6.98e-03, grad_scale: 8.0 +2023-02-06 10:25:22,271 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84682.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:25:30,954 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3011, 1.7133, 1.7306, 1.5565, 1.0600, 1.4520, 1.8830, 1.5660], + device='cuda:0'), covar=tensor([0.0454, 0.1180, 0.1666, 0.1308, 0.0563, 0.1505, 0.0688, 0.0636], + device='cuda:0'), in_proj_covar=tensor([0.0101, 0.0155, 0.0193, 0.0160, 0.0105, 0.0165, 0.0118, 0.0140], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0006], + device='cuda:0') +2023-02-06 10:25:32,800 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=84698.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:25:39,737 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84707.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:25:43,775 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84713.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:25:47,051 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84718.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:25:51,579 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-06 10:25:55,494 INFO [train.py:901] (0/4) Epoch 11, batch 3900, loss[loss=0.2503, simple_loss=0.3213, pruned_loss=0.08966, over 8083.00 frames. ], tot_loss[loss=0.2382, simple_loss=0.3122, pruned_loss=0.08207, over 1608468.41 frames. ], batch size: 21, lr: 6.97e-03, grad_scale: 8.0 +2023-02-06 10:26:03,705 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84743.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:26:08,296 INFO [optim.py:369] (0/4) 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 10:26:30,326 INFO [train.py:901] (0/4) Epoch 11, batch 3950, loss[loss=0.2534, simple_loss=0.3342, pruned_loss=0.08631, over 8283.00 frames. ], tot_loss[loss=0.2367, simple_loss=0.3118, pruned_loss=0.08078, over 1609469.26 frames. ], batch size: 23, lr: 6.97e-03, grad_scale: 8.0 +2023-02-06 10:26:34,004 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.6563, 2.2653, 3.7606, 2.7727, 3.1596, 2.2636, 1.9069, 1.6607], + device='cuda:0'), covar=tensor([0.3525, 0.3928, 0.0952, 0.2371, 0.1839, 0.2089, 0.1628, 0.4391], + device='cuda:0'), in_proj_covar=tensor([0.0882, 0.0864, 0.0729, 0.0835, 0.0932, 0.0794, 0.0703, 0.0762], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 10:26:52,760 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84813.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:26:56,203 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84818.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:27:04,808 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 11, batch 4000, loss[loss=0.2373, simple_loss=0.3244, pruned_loss=0.07504, over 8323.00 frames. ], tot_loss[loss=0.2369, simple_loss=0.3121, pruned_loss=0.08084, over 1612059.96 frames. ], batch size: 25, lr: 6.97e-03, grad_scale: 8.0 +2023-02-06 10:27:17,173 INFO [optim.py:369] (0/4) 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 10:27:21,509 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84855.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:27:33,030 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.5342, 4.4462, 3.9640, 2.1199, 3.9372, 4.0206, 4.1814, 3.7715], + device='cuda:0'), covar=tensor([0.0661, 0.0513, 0.0846, 0.4602, 0.0729, 0.0902, 0.0986, 0.0809], + device='cuda:0'), in_proj_covar=tensor([0.0462, 0.0368, 0.0375, 0.0479, 0.0370, 0.0372, 0.0369, 0.0327], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 10:27:34,470 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.6447, 1.6766, 4.3334, 2.0718, 2.4114, 5.0786, 4.9867, 4.2216], + device='cuda:0'), covar=tensor([0.0968, 0.1718, 0.0312, 0.1842, 0.1140, 0.0162, 0.0351, 0.0623], + device='cuda:0'), in_proj_covar=tensor([0.0264, 0.0300, 0.0259, 0.0292, 0.0272, 0.0239, 0.0340, 0.0294], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 10:27:39,661 INFO [train.py:901] (0/4) Epoch 11, batch 4050, loss[loss=0.2329, simple_loss=0.3035, pruned_loss=0.08114, over 7639.00 frames. ], tot_loss[loss=0.237, simple_loss=0.3121, pruned_loss=0.08095, over 1610902.88 frames. ], batch size: 19, lr: 6.97e-03, grad_scale: 8.0 +2023-02-06 10:27:43,785 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=84887.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:28:07,229 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 +2023-02-06 10:28:15,558 INFO [train.py:901] (0/4) Epoch 11, batch 4100, loss[loss=0.1992, simple_loss=0.2796, pruned_loss=0.0594, over 8090.00 frames. ], tot_loss[loss=0.2357, simple_loss=0.3111, pruned_loss=0.08021, over 1612737.88 frames. ], batch size: 21, lr: 6.97e-03, grad_scale: 8.0 +2023-02-06 10:28:16,473 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84932.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:28:27,741 INFO [optim.py:369] (0/4) 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 10:28:33,379 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84957.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:28:41,349 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84969.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:28:41,430 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84969.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:28:49,615 INFO [train.py:901] (0/4) Epoch 11, batch 4150, loss[loss=0.2599, simple_loss=0.3272, pruned_loss=0.09631, over 8248.00 frames. ], tot_loss[loss=0.2375, simple_loss=0.3126, pruned_loss=0.0812, over 1614025.51 frames. ], batch size: 24, lr: 6.96e-03, grad_scale: 8.0 +2023-02-06 10:28:58,535 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84994.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:28:59,078 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=84995.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:29:04,354 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85002.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:29:11,613 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9971, 3.8820, 2.3073, 2.6271, 2.6639, 1.8203, 2.7594, 2.9552], + device='cuda:0'), covar=tensor([0.1609, 0.0290, 0.0997, 0.0780, 0.0722, 0.1353, 0.1050, 0.1009], + device='cuda:0'), in_proj_covar=tensor([0.0345, 0.0236, 0.0316, 0.0294, 0.0302, 0.0323, 0.0342, 0.0307], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 10:29:23,829 INFO [train.py:901] (0/4) Epoch 11, batch 4200, loss[loss=0.278, simple_loss=0.3503, pruned_loss=0.1029, over 8251.00 frames. ], tot_loss[loss=0.2368, simple_loss=0.3115, pruned_loss=0.08109, over 1611918.71 frames. ], batch size: 24, lr: 6.96e-03, grad_scale: 8.0 +2023-02-06 10:29:36,444 INFO [optim.py:369] (0/4) 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 10:29:47,908 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-06 10:29:50,128 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85069.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:29:53,724 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 +2023-02-06 10:29:58,024 INFO [train.py:901] (0/4) Epoch 11, batch 4250, loss[loss=0.2714, simple_loss=0.3376, pruned_loss=0.1026, over 8523.00 frames. ], tot_loss[loss=0.2364, simple_loss=0.3114, pruned_loss=0.0807, over 1614353.13 frames. ], batch size: 28, lr: 6.96e-03, grad_scale: 8.0 +2023-02-06 10:30:06,835 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85094.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:30:10,052 WARNING [train.py:1067] (0/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-06 10:30:18,058 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85110.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:30:28,605 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85125.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:30:32,495 INFO [train.py:901] (0/4) Epoch 11, batch 4300, loss[loss=0.2143, simple_loss=0.3075, pruned_loss=0.06055, over 8613.00 frames. ], tot_loss[loss=0.2376, simple_loss=0.3124, pruned_loss=0.0814, over 1612991.86 frames. ], batch size: 34, lr: 6.96e-03, grad_scale: 8.0 +2023-02-06 10:30:33,952 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85133.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:30:45,173 INFO [optim.py:369] (0/4) 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 10:30:54,016 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=85162.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:31:06,959 INFO [train.py:901] (0/4) Epoch 11, batch 4350, loss[loss=0.1886, simple_loss=0.2651, pruned_loss=0.05604, over 7538.00 frames. ], tot_loss[loss=0.2377, simple_loss=0.3122, pruned_loss=0.08155, over 1611356.15 frames. ], batch size: 18, lr: 6.96e-03, grad_scale: 8.0 +2023-02-06 10:31:13,424 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 +2023-02-06 10:31:29,831 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6934, 2.0098, 2.2576, 1.2433, 2.3079, 1.4977, 0.7054, 1.9039], + device='cuda:0'), covar=tensor([0.0471, 0.0237, 0.0163, 0.0418, 0.0257, 0.0613, 0.0641, 0.0199], + device='cuda:0'), in_proj_covar=tensor([0.0391, 0.0325, 0.0271, 0.0382, 0.0311, 0.0475, 0.0359, 0.0354], + device='cuda:0'), out_proj_covar=tensor([1.1184e-04, 9.0938e-05, 7.5806e-05, 1.0777e-04, 8.8630e-05, 1.4564e-04, + 1.0287e-04, 1.0083e-04], device='cuda:0') +2023-02-06 10:31:31,294 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 +2023-02-06 10:31:40,295 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 11, batch 4400, loss[loss=0.2969, simple_loss=0.3551, pruned_loss=0.1193, over 8363.00 frames. ], tot_loss[loss=0.2349, simple_loss=0.3097, pruned_loss=0.08004, over 1605080.14 frames. ], batch size: 24, lr: 6.95e-03, grad_scale: 8.0 +2023-02-06 10:31:44,518 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2727, 2.1583, 1.5147, 1.8980, 1.6185, 1.3316, 1.5104, 1.5701], + device='cuda:0'), covar=tensor([0.1203, 0.0369, 0.1142, 0.0564, 0.0775, 0.1397, 0.1053, 0.0792], + device='cuda:0'), in_proj_covar=tensor([0.0348, 0.0237, 0.0318, 0.0296, 0.0301, 0.0324, 0.0342, 0.0308], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 10:31:54,342 INFO [optim.py:369] (0/4) 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 10:32:01,425 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85258.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:32:14,066 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85277.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:32:16,484 INFO [train.py:901] (0/4) Epoch 11, batch 4450, loss[loss=0.2643, simple_loss=0.3303, pruned_loss=0.09918, over 8124.00 frames. ], tot_loss[loss=0.2347, simple_loss=0.3093, pruned_loss=0.08007, over 1601776.28 frames. ], batch size: 22, lr: 6.95e-03, grad_scale: 8.0 +2023-02-06 10:32:18,774 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85283.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:32:22,192 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0673, 2.3304, 1.9517, 2.9247, 1.4587, 1.6318, 1.9542, 2.4847], + device='cuda:0'), covar=tensor([0.0824, 0.0917, 0.1018, 0.0385, 0.1150, 0.1548, 0.1094, 0.0843], + device='cuda:0'), in_proj_covar=tensor([0.0240, 0.0218, 0.0260, 0.0221, 0.0220, 0.0258, 0.0258, 0.0225], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 10:32:22,675 WARNING [train.py:1067] (0/4) 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] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=85313.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:32:39,591 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4147, 2.0651, 3.0719, 2.4443, 2.6489, 2.1645, 1.7586, 1.5140], + device='cuda:0'), covar=tensor([0.3408, 0.3628, 0.1091, 0.2245, 0.1780, 0.1998, 0.1580, 0.3845], + device='cuda:0'), in_proj_covar=tensor([0.0871, 0.0854, 0.0727, 0.0822, 0.0922, 0.0786, 0.0695, 0.0760], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 10:32:50,703 INFO [train.py:901] (0/4) Epoch 11, batch 4500, loss[loss=0.2672, simple_loss=0.3247, pruned_loss=0.1048, over 8137.00 frames. ], tot_loss[loss=0.2329, simple_loss=0.3076, pruned_loss=0.07907, over 1604725.03 frames. ], batch size: 22, lr: 6.95e-03, grad_scale: 8.0 +2023-02-06 10:32:51,285 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 +2023-02-06 10:32:57,950 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.39 vs. limit=5.0 +2023-02-06 10:33:03,399 INFO [optim.py:369] (0/4) 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 10:33:15,846 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-06 10:33:16,046 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85366.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:33:26,537 INFO [train.py:901] (0/4) Epoch 11, batch 4550, loss[loss=0.2357, simple_loss=0.318, pruned_loss=0.07666, over 7919.00 frames. ], tot_loss[loss=0.2357, simple_loss=0.3105, pruned_loss=0.08051, over 1608267.27 frames. ], batch size: 20, lr: 6.95e-03, grad_scale: 8.0 +2023-02-06 10:33:33,510 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85391.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:33:43,813 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4050, 2.5262, 1.7865, 2.1439, 2.0203, 1.4999, 1.9208, 1.9486], + device='cuda:0'), covar=tensor([0.1160, 0.0324, 0.1010, 0.0522, 0.0615, 0.1288, 0.0824, 0.0801], + device='cuda:0'), in_proj_covar=tensor([0.0353, 0.0240, 0.0322, 0.0300, 0.0307, 0.0328, 0.0347, 0.0312], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 10:33:47,180 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([0.9807, 1.6776, 1.3959, 1.6015, 1.3904, 1.2169, 1.3133, 1.3100], + device='cuda:0'), covar=tensor([0.1015, 0.0452, 0.1072, 0.0503, 0.0620, 0.1339, 0.0769, 0.0731], + device='cuda:0'), in_proj_covar=tensor([0.0352, 0.0239, 0.0321, 0.0299, 0.0306, 0.0327, 0.0346, 0.0311], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 10:33:59,380 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85428.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:34:01,129 INFO [train.py:901] (0/4) Epoch 11, batch 4600, loss[loss=0.2007, simple_loss=0.2781, pruned_loss=0.06162, over 7709.00 frames. ], tot_loss[loss=0.2359, simple_loss=0.3105, pruned_loss=0.08066, over 1609691.18 frames. ], batch size: 18, lr: 6.95e-03, grad_scale: 8.0 +2023-02-06 10:34:11,933 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 10:34:27,368 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=85469.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:34:33,586 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=85477.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:34:36,253 INFO [train.py:901] (0/4) Epoch 11, batch 4650, loss[loss=0.3013, simple_loss=0.3599, pruned_loss=0.1214, over 7195.00 frames. ], tot_loss[loss=0.2357, simple_loss=0.3099, pruned_loss=0.08076, over 1607205.74 frames. ], batch size: 72, lr: 6.94e-03, grad_scale: 8.0 +2023-02-06 10:34:50,386 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85501.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:35:11,112 INFO [train.py:901] (0/4) Epoch 11, batch 4700, loss[loss=0.1949, simple_loss=0.2747, pruned_loss=0.05749, over 7547.00 frames. ], tot_loss[loss=0.2362, simple_loss=0.3106, pruned_loss=0.08093, over 1603933.34 frames. ], batch size: 18, lr: 6.94e-03, grad_scale: 8.0 +2023-02-06 10:35:12,705 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85533.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:35:19,223 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.1971, 1.9250, 2.0328, 1.8869, 1.3258, 1.9237, 2.6764, 2.7243], + device='cuda:0'), covar=tensor([0.0405, 0.1155, 0.1576, 0.1245, 0.0547, 0.1343, 0.0552, 0.0478], + device='cuda:0'), in_proj_covar=tensor([0.0099, 0.0152, 0.0192, 0.0157, 0.0103, 0.0162, 0.0117, 0.0139], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0006], + device='cuda:0') +2023-02-06 10:35:22,553 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85548.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:35:23,046 INFO [optim.py:369] (0/4) 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 10:35:30,010 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85558.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:35:32,597 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7517, 1.7828, 3.3514, 1.4753, 2.2047, 3.5940, 3.6850, 3.0953], + device='cuda:0'), covar=tensor([0.1045, 0.1377, 0.0302, 0.1861, 0.0842, 0.0217, 0.0403, 0.0548], + device='cuda:0'), in_proj_covar=tensor([0.0259, 0.0293, 0.0254, 0.0286, 0.0267, 0.0233, 0.0334, 0.0287], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-06 10:35:39,375 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5408, 1.9716, 3.4109, 1.2540, 2.4659, 1.9012, 1.6131, 2.3666], + device='cuda:0'), covar=tensor([0.1552, 0.2044, 0.0619, 0.3616, 0.1421, 0.2655, 0.1746, 0.1946], + device='cuda:0'), in_proj_covar=tensor([0.0487, 0.0517, 0.0531, 0.0578, 0.0617, 0.0554, 0.0473, 0.0610], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 10:35:45,762 INFO [train.py:901] (0/4) Epoch 11, batch 4750, loss[loss=0.2543, simple_loss=0.325, pruned_loss=0.0918, over 8098.00 frames. ], tot_loss[loss=0.2371, simple_loss=0.3115, pruned_loss=0.08136, over 1605341.83 frames. ], batch size: 23, lr: 6.94e-03, grad_scale: 8.0 +2023-02-06 10:35:47,962 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85584.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:35:53,285 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85592.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:36:06,687 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3127, 1.2814, 1.5125, 1.2553, 0.7147, 1.3301, 1.1872, 1.3171], + device='cuda:0'), covar=tensor([0.0503, 0.1242, 0.1719, 0.1362, 0.0549, 0.1472, 0.0659, 0.0579], + device='cuda:0'), in_proj_covar=tensor([0.0099, 0.0152, 0.0191, 0.0157, 0.0103, 0.0162, 0.0116, 0.0138], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0006], + device='cuda:0') +2023-02-06 10:36:09,841 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-06 10:36:11,811 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-06 10:36:20,655 INFO [train.py:901] (0/4) Epoch 11, batch 4800, loss[loss=0.1841, simple_loss=0.2737, pruned_loss=0.04721, over 7917.00 frames. ], tot_loss[loss=0.2374, simple_loss=0.3116, pruned_loss=0.08162, over 1607444.37 frames. ], batch size: 20, lr: 6.94e-03, grad_scale: 8.0 +2023-02-06 10:36:28,782 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 10:36:55,319 INFO [train.py:901] (0/4) Epoch 11, batch 4850, loss[loss=0.2273, simple_loss=0.3028, pruned_loss=0.07588, over 8242.00 frames. ], tot_loss[loss=0.236, simple_loss=0.3107, pruned_loss=0.08062, over 1612471.79 frames. ], batch size: 22, lr: 6.94e-03, grad_scale: 8.0 +2023-02-06 10:36:57,577 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85684.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:37:01,303 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-06 10:37:12,690 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4133, 1.7184, 2.6634, 1.2178, 2.0968, 1.7473, 1.5524, 1.8679], + device='cuda:0'), covar=tensor([0.1799, 0.2412, 0.0771, 0.4021, 0.1540, 0.2850, 0.1877, 0.2084], + device='cuda:0'), in_proj_covar=tensor([0.0490, 0.0521, 0.0535, 0.0584, 0.0621, 0.0555, 0.0474, 0.0615], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 10:37:13,008 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 +2023-02-06 10:37:14,634 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85709.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:37:17,363 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5738, 1.6793, 3.2575, 1.2189, 2.1255, 3.5505, 3.5862, 2.9809], + device='cuda:0'), covar=tensor([0.1236, 0.1449, 0.0357, 0.2292, 0.1015, 0.0262, 0.0542, 0.0678], + device='cuda:0'), in_proj_covar=tensor([0.0259, 0.0293, 0.0256, 0.0289, 0.0267, 0.0233, 0.0335, 0.0287], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-06 10:37:30,204 INFO [train.py:901] (0/4) Epoch 11, batch 4900, loss[loss=0.2502, simple_loss=0.318, pruned_loss=0.09119, over 8080.00 frames. ], tot_loss[loss=0.2364, simple_loss=0.3112, pruned_loss=0.08079, over 1615486.87 frames. ], batch size: 21, lr: 6.93e-03, grad_scale: 8.0 +2023-02-06 10:37:42,891 INFO [optim.py:369] (0/4) 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 10:37:58,837 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7309, 1.7782, 2.1428, 1.5875, 1.1585, 2.2035, 0.2063, 1.2530], + device='cuda:0'), covar=tensor([0.2595, 0.1635, 0.0517, 0.1982, 0.4619, 0.0508, 0.3485, 0.2073], + device='cuda:0'), in_proj_covar=tensor([0.0164, 0.0165, 0.0098, 0.0212, 0.0254, 0.0103, 0.0165, 0.0164], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 10:38:04,658 INFO [train.py:901] (0/4) Epoch 11, batch 4950, loss[loss=0.2522, simple_loss=0.3273, pruned_loss=0.08857, over 8140.00 frames. ], tot_loss[loss=0.2364, simple_loss=0.3114, pruned_loss=0.08071, over 1615937.75 frames. ], batch size: 22, lr: 6.93e-03, grad_scale: 8.0 +2023-02-06 10:38:11,375 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=85790.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:38:36,552 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3868, 1.5938, 2.5286, 0.9700, 1.9685, 2.7819, 3.0280, 1.9780], + device='cuda:0'), covar=tensor([0.1549, 0.1818, 0.0710, 0.3109, 0.1121, 0.0571, 0.0775, 0.1500], + device='cuda:0'), in_proj_covar=tensor([0.0258, 0.0293, 0.0255, 0.0286, 0.0266, 0.0231, 0.0334, 0.0287], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-06 10:38:39,604 INFO [train.py:901] (0/4) Epoch 11, batch 5000, loss[loss=0.2214, simple_loss=0.3111, pruned_loss=0.06585, over 8285.00 frames. ], tot_loss[loss=0.2375, simple_loss=0.3124, pruned_loss=0.0813, over 1615617.66 frames. ], batch size: 23, lr: 6.93e-03, grad_scale: 8.0 +2023-02-06 10:38:46,503 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85840.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:38:47,419 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.76 vs. limit=5.0 +2023-02-06 10:38:49,800 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=85845.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:38:51,895 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 10:39:03,553 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85865.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:39:08,900 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85873.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:39:13,944 INFO [train.py:901] (0/4) Epoch 11, batch 5050, loss[loss=0.2852, simple_loss=0.3455, pruned_loss=0.1124, over 6866.00 frames. ], tot_loss[loss=0.2391, simple_loss=0.3134, pruned_loss=0.08239, over 1612949.63 frames. ], batch size: 71, lr: 6.93e-03, grad_scale: 8.0 +2023-02-06 10:39:15,552 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.40 vs. limit=5.0 +2023-02-06 10:39:21,248 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=85892.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:39:22,000 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85893.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:39:30,131 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85905.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:39:39,822 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-06 10:39:48,589 INFO [train.py:901] (0/4) Epoch 11, batch 5100, loss[loss=0.246, simple_loss=0.3192, pruned_loss=0.08645, over 7984.00 frames. ], tot_loss[loss=0.2367, simple_loss=0.3116, pruned_loss=0.08088, over 1608606.37 frames. ], batch size: 21, lr: 6.93e-03, grad_scale: 8.0 +2023-02-06 10:40:00,784 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85948.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 10:40:01,254 INFO [optim.py:369] (0/4) 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 10:40:08,917 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85960.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:40:19,542 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85975.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:40:23,495 INFO [train.py:901] (0/4) Epoch 11, batch 5150, loss[loss=0.2989, simple_loss=0.3628, pruned_loss=0.1175, over 8626.00 frames. ], tot_loss[loss=0.237, simple_loss=0.312, pruned_loss=0.08098, over 1609774.71 frames. ], batch size: 34, lr: 6.92e-03, grad_scale: 8.0 +2023-02-06 10:40:27,616 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=85987.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:40:36,305 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-86000.pt +2023-02-06 10:40:42,287 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86007.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:40:51,412 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86020.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:40:59,442 INFO [train.py:901] (0/4) Epoch 11, batch 5200, loss[loss=0.2747, simple_loss=0.3412, pruned_loss=0.1041, over 8262.00 frames. ], tot_loss[loss=0.2362, simple_loss=0.3117, pruned_loss=0.08035, over 1611499.24 frames. ], batch size: 49, lr: 6.92e-03, grad_scale: 16.0 +2023-02-06 10:41:12,341 INFO [optim.py:369] (0/4) 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 10:41:22,283 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8390, 1.5006, 3.1341, 1.3701, 2.0604, 3.4171, 3.4175, 2.9331], + device='cuda:0'), covar=tensor([0.1039, 0.1476, 0.0340, 0.1878, 0.0893, 0.0262, 0.0515, 0.0622], + device='cuda:0'), in_proj_covar=tensor([0.0263, 0.0297, 0.0258, 0.0291, 0.0272, 0.0236, 0.0339, 0.0293], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 10:41:27,748 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86070.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:41:35,337 INFO [train.py:901] (0/4) Epoch 11, batch 5250, loss[loss=0.3236, simple_loss=0.3645, pruned_loss=0.1413, over 6567.00 frames. ], tot_loss[loss=0.2369, simple_loss=0.3119, pruned_loss=0.08094, over 1607526.01 frames. ], batch size: 71, lr: 6.92e-03, grad_scale: 16.0 +2023-02-06 10:41:39,733 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86087.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:41:40,973 WARNING [train.py:1067] (0/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-06 10:41:50,767 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86102.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:42:10,648 INFO [train.py:901] (0/4) Epoch 11, batch 5300, loss[loss=0.2355, simple_loss=0.3055, pruned_loss=0.08279, over 7802.00 frames. ], tot_loss[loss=0.2359, simple_loss=0.3111, pruned_loss=0.08033, over 1607052.31 frames. ], batch size: 20, lr: 6.92e-03, grad_scale: 16.0 +2023-02-06 10:42:14,346 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7763, 2.3277, 4.4932, 1.4700, 3.2781, 2.2700, 1.7300, 2.9574], + device='cuda:0'), covar=tensor([0.1640, 0.2144, 0.0654, 0.3821, 0.1453, 0.2744, 0.1858, 0.2326], + device='cuda:0'), in_proj_covar=tensor([0.0485, 0.0519, 0.0533, 0.0580, 0.0614, 0.0551, 0.0471, 0.0611], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 10:42:23,760 INFO [optim.py:369] (0/4) 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 10:42:32,860 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86161.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:42:46,503 INFO [train.py:901] (0/4) Epoch 11, batch 5350, loss[loss=0.2432, simple_loss=0.3161, pruned_loss=0.08514, over 8443.00 frames. ], tot_loss[loss=0.2364, simple_loss=0.3115, pruned_loss=0.08066, over 1606722.98 frames. ], batch size: 27, lr: 6.92e-03, grad_scale: 16.0 +2023-02-06 10:42:50,718 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86186.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:43:12,446 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86216.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:43:22,285 INFO [train.py:901] (0/4) Epoch 11, batch 5400, loss[loss=0.258, simple_loss=0.3293, pruned_loss=0.09329, over 8074.00 frames. ], tot_loss[loss=0.2364, simple_loss=0.3113, pruned_loss=0.08073, over 1607738.24 frames. ], batch size: 21, lr: 6.91e-03, grad_scale: 16.0 +2023-02-06 10:43:26,581 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=86237.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:43:29,483 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86241.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:43:34,239 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5581, 1.5487, 1.8679, 1.6157, 1.1685, 1.9524, 0.4103, 1.3039], + device='cuda:0'), covar=tensor([0.2333, 0.1454, 0.0625, 0.1388, 0.4001, 0.0418, 0.2838, 0.1844], + device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0164, 0.0096, 0.0208, 0.0246, 0.0100, 0.0160, 0.0161], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 10:43:34,674 INFO [optim.py:369] (0/4) 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 10:43:44,521 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86263.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:43:45,145 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86264.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:43:57,388 INFO [train.py:901] (0/4) Epoch 11, batch 5450, loss[loss=0.2388, simple_loss=0.3037, pruned_loss=0.08698, over 7422.00 frames. ], tot_loss[loss=0.2369, simple_loss=0.3115, pruned_loss=0.08119, over 1607401.21 frames. ], batch size: 17, lr: 6.91e-03, grad_scale: 16.0 +2023-02-06 10:44:03,050 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86288.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:44:05,805 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=86292.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 10:44:07,248 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6690, 1.8009, 1.4855, 2.2306, 1.0665, 1.3929, 1.5818, 1.8885], + device='cuda:0'), covar=tensor([0.0811, 0.0895, 0.1111, 0.0519, 0.1185, 0.1407, 0.0909, 0.0770], + device='cuda:0'), in_proj_covar=tensor([0.0238, 0.0215, 0.0254, 0.0216, 0.0217, 0.0253, 0.0254, 0.0220], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 10:44:25,120 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=86319.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:44:34,079 INFO [train.py:901] (0/4) Epoch 11, batch 5500, loss[loss=0.2336, simple_loss=0.3263, pruned_loss=0.07043, over 8326.00 frames. ], tot_loss[loss=0.2363, simple_loss=0.3116, pruned_loss=0.08049, over 1612184.60 frames. ], batch size: 25, lr: 6.91e-03, grad_scale: 16.0 +2023-02-06 10:44:34,721 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-06 10:44:44,468 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 +2023-02-06 10:44:46,144 INFO [optim.py:369] (0/4) 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 10:44:48,449 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86352.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:44:52,684 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86358.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:44:56,767 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=86364.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:45:04,260 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 +2023-02-06 10:45:09,173 INFO [train.py:901] (0/4) Epoch 11, batch 5550, loss[loss=0.2994, simple_loss=0.3617, pruned_loss=0.1186, over 8594.00 frames. ], tot_loss[loss=0.2365, simple_loss=0.3116, pruned_loss=0.08072, over 1614176.44 frames. ], batch size: 39, lr: 6.91e-03, grad_scale: 16.0 +2023-02-06 10:45:10,769 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86383.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:45:27,853 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86407.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 10:45:33,155 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=86414.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:45:44,330 INFO [train.py:901] (0/4) Epoch 11, batch 5600, loss[loss=0.2505, simple_loss=0.3405, pruned_loss=0.08028, over 8333.00 frames. ], tot_loss[loss=0.2366, simple_loss=0.3119, pruned_loss=0.08067, over 1614572.58 frames. ], batch size: 25, lr: 6.91e-03, grad_scale: 16.0 +2023-02-06 10:45:44,398 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=86431.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:45:46,515 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86434.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:45:57,251 INFO [optim.py:369] (0/4) 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 10:46:03,244 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86458.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:46:06,583 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86463.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:46:17,331 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86479.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:46:18,549 INFO [train.py:901] (0/4) Epoch 11, batch 5650, loss[loss=0.2572, simple_loss=0.3405, pruned_loss=0.08691, over 8459.00 frames. ], tot_loss[loss=0.2379, simple_loss=0.3129, pruned_loss=0.08147, over 1614247.75 frames. ], batch size: 25, lr: 6.90e-03, grad_scale: 16.0 +2023-02-06 10:46:37,004 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 +2023-02-06 10:46:39,906 WARNING [train.py:1067] (0/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-06 10:46:44,006 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.1427, 1.1848, 3.3442, 0.8983, 2.9275, 2.8315, 3.0671, 2.9537], + device='cuda:0'), covar=tensor([0.0906, 0.3870, 0.0808, 0.3813, 0.1605, 0.1031, 0.0791, 0.0915], + device='cuda:0'), in_proj_covar=tensor([0.0491, 0.0576, 0.0579, 0.0534, 0.0605, 0.0517, 0.0509, 0.0582], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 10:46:52,193 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86529.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:46:53,988 INFO [train.py:901] (0/4) Epoch 11, batch 5700, loss[loss=0.2131, simple_loss=0.2802, pruned_loss=0.07299, over 7810.00 frames. ], tot_loss[loss=0.2361, simple_loss=0.3113, pruned_loss=0.08045, over 1614207.92 frames. ], batch size: 20, lr: 6.90e-03, grad_scale: 16.0 +2023-02-06 10:47:04,198 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 10:47:11,018 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8068, 1.6624, 3.4727, 1.4811, 2.3862, 3.8695, 3.8838, 3.2953], + device='cuda:0'), covar=tensor([0.1173, 0.1547, 0.0335, 0.1988, 0.0968, 0.0182, 0.0390, 0.0562], + device='cuda:0'), in_proj_covar=tensor([0.0260, 0.0295, 0.0255, 0.0286, 0.0268, 0.0233, 0.0335, 0.0287], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-06 10:47:28,594 INFO [train.py:901] (0/4) Epoch 11, batch 5750, loss[loss=0.2034, simple_loss=0.3019, pruned_loss=0.05242, over 8317.00 frames. ], tot_loss[loss=0.2349, simple_loss=0.3102, pruned_loss=0.07984, over 1610664.19 frames. ], batch size: 25, lr: 6.90e-03, grad_scale: 16.0 +2023-02-06 10:47:29,098 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.33 vs. limit=5.0 +2023-02-06 10:47:37,014 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 +2023-02-06 10:47:40,219 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86598.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:47:42,138 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-06 10:47:47,612 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=86608.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:47:47,756 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86608.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:48:03,810 INFO [train.py:901] (0/4) Epoch 11, batch 5800, loss[loss=0.2397, simple_loss=0.3202, pruned_loss=0.07956, over 8568.00 frames. ], tot_loss[loss=0.2364, simple_loss=0.3114, pruned_loss=0.08068, over 1610658.97 frames. ], batch size: 31, lr: 6.90e-03, grad_scale: 16.0 +2023-02-06 10:48:05,399 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86633.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:48:17,064 INFO [optim.py:369] (0/4) 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 10:48:19,991 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9456, 3.4153, 2.2436, 2.3880, 2.4618, 2.0112, 2.4986, 2.7344], + device='cuda:0'), covar=tensor([0.1441, 0.0247, 0.0926, 0.0777, 0.0674, 0.1170, 0.0917, 0.1012], + device='cuda:0'), in_proj_covar=tensor([0.0345, 0.0231, 0.0311, 0.0293, 0.0300, 0.0317, 0.0337, 0.0301], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 10:48:26,848 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 11, batch 5850, loss[loss=0.2602, simple_loss=0.3363, pruned_loss=0.09211, over 8091.00 frames. ], tot_loss[loss=0.2344, simple_loss=0.3099, pruned_loss=0.07942, over 1610303.89 frames. ], batch size: 21, lr: 6.90e-03, grad_scale: 16.0 +2023-02-06 10:48:44,276 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86688.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 10:48:45,640 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86690.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:49:02,114 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86715.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:49:02,974 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 +2023-02-06 10:49:07,935 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86723.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:49:13,225 INFO [train.py:901] (0/4) Epoch 11, batch 5900, loss[loss=0.2489, simple_loss=0.3221, pruned_loss=0.08785, over 6942.00 frames. ], tot_loss[loss=0.2363, simple_loss=0.3118, pruned_loss=0.0804, over 1611781.51 frames. ], batch size: 71, lr: 6.89e-03, grad_scale: 16.0 +2023-02-06 10:49:16,683 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86735.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:49:25,721 INFO [optim.py:369] (0/4) 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 10:49:33,349 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86760.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:49:43,863 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=2.12 vs. limit=2.0 +2023-02-06 10:49:48,264 INFO [train.py:901] (0/4) Epoch 11, batch 5950, loss[loss=0.2003, simple_loss=0.2779, pruned_loss=0.06136, over 7809.00 frames. ], tot_loss[loss=0.2364, simple_loss=0.3116, pruned_loss=0.08063, over 1614783.80 frames. ], batch size: 20, lr: 6.89e-03, grad_scale: 16.0 +2023-02-06 10:49:51,961 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86785.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:49:55,662 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 +2023-02-06 10:50:03,637 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=86807.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:50:09,056 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86810.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:50:20,470 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86827.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:50:22,926 INFO [train.py:901] (0/4) Epoch 11, batch 6000, loss[loss=0.216, simple_loss=0.2923, pruned_loss=0.06988, over 8252.00 frames. ], tot_loss[loss=0.2358, simple_loss=0.3111, pruned_loss=0.08021, over 1614747.07 frames. ], batch size: 22, lr: 6.89e-03, grad_scale: 16.0 +2023-02-06 10:50:22,927 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 10:50:35,336 INFO [train.py:935] (0/4) Epoch 11, validation: loss=0.1887, simple_loss=0.2887, pruned_loss=0.04439, over 944034.00 frames. +2023-02-06 10:50:35,340 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-06 10:50:36,202 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 10:51:10,313 INFO [train.py:901] (0/4) Epoch 11, batch 6050, loss[loss=0.2177, simple_loss=0.2998, pruned_loss=0.06777, over 8245.00 frames. ], tot_loss[loss=0.2361, simple_loss=0.3117, pruned_loss=0.08022, over 1619515.09 frames. ], batch size: 22, lr: 6.89e-03, grad_scale: 16.0 +2023-02-06 10:51:20,508 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86896.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:51:35,565 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86917.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:51:39,056 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86922.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:51:44,785 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.5997, 2.2009, 3.4916, 2.4386, 2.9417, 2.4019, 1.9349, 1.7977], + device='cuda:0'), covar=tensor([0.4267, 0.4621, 0.1229, 0.3382, 0.2491, 0.2152, 0.1598, 0.4540], + device='cuda:0'), in_proj_covar=tensor([0.0884, 0.0867, 0.0727, 0.0837, 0.0929, 0.0792, 0.0699, 0.0762], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 10:51:45,217 INFO [train.py:901] (0/4) Epoch 11, batch 6100, loss[loss=0.1704, simple_loss=0.263, pruned_loss=0.03894, over 8035.00 frames. ], tot_loss[loss=0.2337, simple_loss=0.3096, pruned_loss=0.0789, over 1614157.08 frames. ], batch size: 22, lr: 6.89e-03, grad_scale: 16.0 +2023-02-06 10:51:46,288 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=2.22 vs. limit=2.0 +2023-02-06 10:51:53,476 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=86942.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:51:58,237 INFO [optim.py:369] (0/4) 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 10:52:06,848 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-06 10:52:19,156 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86979.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:52:20,326 INFO [train.py:901] (0/4) Epoch 11, batch 6150, loss[loss=0.2133, simple_loss=0.2944, pruned_loss=0.06606, over 8141.00 frames. ], tot_loss[loss=0.2347, simple_loss=0.3103, pruned_loss=0.07952, over 1616400.71 frames. ], batch size: 22, lr: 6.88e-03, grad_scale: 16.0 +2023-02-06 10:52:36,914 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87004.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:52:47,749 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 11, batch 6200, loss[loss=0.2617, simple_loss=0.3303, pruned_loss=0.09657, over 7056.00 frames. ], tot_loss[loss=0.2339, simple_loss=0.3097, pruned_loss=0.07907, over 1612110.59 frames. ], batch size: 71, lr: 6.88e-03, grad_scale: 16.0 +2023-02-06 10:53:00,829 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6427, 1.6611, 2.0028, 1.5717, 1.0820, 2.0030, 0.2977, 1.3147], + device='cuda:0'), covar=tensor([0.2884, 0.1527, 0.0595, 0.1739, 0.4426, 0.0517, 0.3507, 0.1875], + device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0167, 0.0099, 0.0212, 0.0251, 0.0103, 0.0163, 0.0164], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 10:53:07,897 INFO [optim.py:369] (0/4) 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 10:53:14,420 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87057.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:53:18,218 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 +2023-02-06 10:53:30,916 INFO [train.py:901] (0/4) Epoch 11, batch 6250, loss[loss=0.1978, simple_loss=0.2682, pruned_loss=0.06369, over 7703.00 frames. ], tot_loss[loss=0.2349, simple_loss=0.3104, pruned_loss=0.07969, over 1609031.45 frames. ], batch size: 18, lr: 6.88e-03, grad_scale: 16.0 +2023-02-06 10:53:39,949 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([6.0536, 6.1462, 5.3280, 2.4376, 5.3927, 5.7252, 5.7789, 5.4326], + device='cuda:0'), covar=tensor([0.0534, 0.0437, 0.0803, 0.4145, 0.0602, 0.0520, 0.1022, 0.0513], + device='cuda:0'), in_proj_covar=tensor([0.0463, 0.0369, 0.0382, 0.0478, 0.0375, 0.0375, 0.0373, 0.0332], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 10:54:06,553 INFO [train.py:901] (0/4) Epoch 11, batch 6300, loss[loss=0.2992, simple_loss=0.3533, pruned_loss=0.1226, over 6669.00 frames. ], tot_loss[loss=0.2361, simple_loss=0.3118, pruned_loss=0.08024, over 1610598.59 frames. ], batch size: 71, lr: 6.88e-03, grad_scale: 16.0 +2023-02-06 10:54:19,292 INFO [optim.py:369] (0/4) 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 10:54:23,472 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.8052, 3.6970, 3.3866, 1.8320, 3.3328, 3.2726, 3.3920, 3.0482], + device='cuda:0'), covar=tensor([0.0965, 0.0752, 0.1155, 0.4710, 0.1009, 0.1121, 0.1464, 0.1106], + device='cuda:0'), in_proj_covar=tensor([0.0461, 0.0367, 0.0380, 0.0477, 0.0375, 0.0372, 0.0371, 0.0330], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 10:54:36,427 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87173.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:54:38,304 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=87176.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:54:39,796 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87178.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:54:41,628 INFO [train.py:901] (0/4) Epoch 11, batch 6350, loss[loss=0.2552, simple_loss=0.3197, pruned_loss=0.09536, over 8128.00 frames. ], tot_loss[loss=0.2361, simple_loss=0.3119, pruned_loss=0.08014, over 1614235.49 frames. ], batch size: 22, lr: 6.88e-03, grad_scale: 16.0 +2023-02-06 10:54:53,183 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87198.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:54:57,269 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 11, batch 6400, loss[loss=0.2476, simple_loss=0.3342, pruned_loss=0.0805, over 8526.00 frames. ], tot_loss[loss=0.2362, simple_loss=0.3123, pruned_loss=0.08006, over 1619783.92 frames. ], batch size: 26, lr: 6.87e-03, grad_scale: 16.0 +2023-02-06 10:55:19,098 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.1631, 1.8188, 1.3802, 1.5754, 1.5331, 1.2178, 1.4531, 1.4814], + device='cuda:0'), covar=tensor([0.0873, 0.0266, 0.0738, 0.0421, 0.0474, 0.0916, 0.0608, 0.0597], + device='cuda:0'), in_proj_covar=tensor([0.0347, 0.0235, 0.0315, 0.0296, 0.0300, 0.0319, 0.0340, 0.0304], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 10:55:23,187 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=87240.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:55:28,875 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87248.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:55:29,366 INFO [optim.py:369] (0/4) 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 10:55:51,533 INFO [train.py:901] (0/4) Epoch 11, batch 6450, loss[loss=0.1989, simple_loss=0.2732, pruned_loss=0.06233, over 7707.00 frames. ], tot_loss[loss=0.2366, simple_loss=0.3121, pruned_loss=0.08055, over 1614483.33 frames. ], batch size: 18, lr: 6.87e-03, grad_scale: 16.0 +2023-02-06 10:55:59,186 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87291.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:56:14,145 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87313.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:56:27,326 INFO [train.py:901] (0/4) Epoch 11, batch 6500, loss[loss=0.2363, simple_loss=0.3125, pruned_loss=0.08005, over 6790.00 frames. ], tot_loss[loss=0.2356, simple_loss=0.3107, pruned_loss=0.08023, over 1610659.73 frames. ], batch size: 74, lr: 6.87e-03, grad_scale: 16.0 +2023-02-06 10:56:32,327 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87338.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:56:39,860 INFO [optim.py:369] (0/4) 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 10:56:44,213 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87355.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:56:50,401 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=87364.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 10:56:58,811 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-02-06 10:57:01,880 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87380.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:57:02,427 INFO [train.py:901] (0/4) Epoch 11, batch 6550, loss[loss=0.2184, simple_loss=0.3003, pruned_loss=0.06826, over 8033.00 frames. ], tot_loss[loss=0.2346, simple_loss=0.31, pruned_loss=0.07965, over 1608838.29 frames. ], batch size: 22, lr: 6.87e-03, grad_scale: 16.0 +2023-02-06 10:57:04,032 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1357, 1.4941, 1.5585, 1.2341, 0.9076, 1.3188, 1.7512, 1.4513], + device='cuda:0'), covar=tensor([0.0516, 0.1319, 0.1857, 0.1524, 0.0648, 0.1661, 0.0757, 0.0684], + device='cuda:0'), in_proj_covar=tensor([0.0100, 0.0153, 0.0192, 0.0159, 0.0104, 0.0165, 0.0117, 0.0138], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0006], + device='cuda:0') +2023-02-06 10:57:17,755 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-06 10:57:37,067 INFO [train.py:901] (0/4) Epoch 11, batch 6600, loss[loss=0.2313, simple_loss=0.3129, pruned_loss=0.07487, over 8129.00 frames. ], tot_loss[loss=0.2344, simple_loss=0.3102, pruned_loss=0.07931, over 1613551.52 frames. ], batch size: 22, lr: 6.87e-03, grad_scale: 16.0 +2023-02-06 10:57:37,792 WARNING [train.py:1067] (0/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-06 10:57:39,954 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5628, 1.5259, 1.7758, 1.4259, 1.0538, 1.8386, 0.1008, 1.1624], + device='cuda:0'), covar=tensor([0.2450, 0.1795, 0.0722, 0.1800, 0.4472, 0.0591, 0.3170, 0.2151], + device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0166, 0.0099, 0.0213, 0.0250, 0.0101, 0.0160, 0.0163], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 10:57:50,090 INFO [optim.py:369] (0/4) 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 10:57:58,511 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6508, 1.9523, 3.0446, 1.4199, 2.2195, 2.0694, 1.6596, 1.9822], + device='cuda:0'), covar=tensor([0.1689, 0.2289, 0.0792, 0.3858, 0.1527, 0.2786, 0.1838, 0.2314], + device='cuda:0'), in_proj_covar=tensor([0.0481, 0.0517, 0.0530, 0.0573, 0.0610, 0.0549, 0.0466, 0.0607], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 10:58:11,488 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87479.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:58:11,525 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 11, batch 6650, loss[loss=0.2226, simple_loss=0.3125, pruned_loss=0.0663, over 8669.00 frames. ], tot_loss[loss=0.234, simple_loss=0.3098, pruned_loss=0.07911, over 1612070.34 frames. ], batch size: 34, lr: 6.86e-03, grad_scale: 16.0 +2023-02-06 10:58:41,507 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87523.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:58:47,483 INFO [train.py:901] (0/4) Epoch 11, batch 6700, loss[loss=0.2249, simple_loss=0.3025, pruned_loss=0.07366, over 7226.00 frames. ], tot_loss[loss=0.2355, simple_loss=0.3108, pruned_loss=0.08005, over 1610033.36 frames. ], batch size: 16, lr: 6.86e-03, grad_scale: 16.0 +2023-02-06 10:58:58,339 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87547.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:58:59,467 INFO [optim.py:369] (0/4) 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 10:59:04,537 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.7440, 1.4020, 5.7920, 2.1050, 5.1970, 4.8972, 5.3716, 5.2226], + device='cuda:0'), covar=tensor([0.0443, 0.4309, 0.0335, 0.3200, 0.0946, 0.0713, 0.0460, 0.0517], + device='cuda:0'), in_proj_covar=tensor([0.0485, 0.0570, 0.0580, 0.0530, 0.0600, 0.0515, 0.0505, 0.0573], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 10:59:16,928 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87572.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:59:18,335 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8788, 1.8660, 2.4204, 1.7868, 1.2698, 2.6279, 0.4463, 1.4845], + device='cuda:0'), covar=tensor([0.2349, 0.1690, 0.0432, 0.1956, 0.3766, 0.0305, 0.3521, 0.2127], + device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0165, 0.0099, 0.0211, 0.0247, 0.0100, 0.0159, 0.0161], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 10:59:18,586 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 +2023-02-06 10:59:22,974 INFO [train.py:901] (0/4) Epoch 11, batch 6750, loss[loss=0.2137, simple_loss=0.2973, pruned_loss=0.06505, over 8333.00 frames. ], tot_loss[loss=0.2352, simple_loss=0.311, pruned_loss=0.07973, over 1614340.60 frames. ], batch size: 25, lr: 6.86e-03, grad_scale: 16.0 +2023-02-06 10:59:27,210 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9590, 1.9911, 1.6514, 2.5412, 1.1033, 1.4901, 1.6379, 2.1427], + device='cuda:0'), covar=tensor([0.0742, 0.0905, 0.1206, 0.0463, 0.1360, 0.1541, 0.1038, 0.0786], + device='cuda:0'), in_proj_covar=tensor([0.0235, 0.0214, 0.0257, 0.0218, 0.0218, 0.0254, 0.0256, 0.0221], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 10:59:30,567 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=87592.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:59:37,569 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87602.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:59:40,793 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87607.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:59:43,433 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87611.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:59:52,274 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87623.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 10:59:56,912 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-06 10:59:57,600 INFO [train.py:901] (0/4) Epoch 11, batch 6800, loss[loss=0.2031, simple_loss=0.2725, pruned_loss=0.06679, over 7791.00 frames. ], tot_loss[loss=0.2355, simple_loss=0.3108, pruned_loss=0.08007, over 1612571.74 frames. ], batch size: 19, lr: 6.86e-03, grad_scale: 16.0 +2023-02-06 11:00:01,247 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87636.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:00:10,517 INFO [optim.py:369] (0/4) 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 11:00:32,377 INFO [train.py:901] (0/4) Epoch 11, batch 6850, loss[loss=0.2551, simple_loss=0.3314, pruned_loss=0.08947, over 8289.00 frames. ], tot_loss[loss=0.2351, simple_loss=0.3105, pruned_loss=0.07981, over 1612091.16 frames. ], batch size: 23, lr: 6.86e-03, grad_scale: 16.0 +2023-02-06 11:00:45,132 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-06 11:00:50,749 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87707.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:00:54,104 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1816, 2.7937, 3.0636, 1.4240, 3.2287, 2.0828, 1.5687, 2.1724], + device='cuda:0'), covar=tensor([0.0608, 0.0250, 0.0229, 0.0559, 0.0359, 0.0583, 0.0674, 0.0334], + device='cuda:0'), in_proj_covar=tensor([0.0391, 0.0327, 0.0269, 0.0384, 0.0312, 0.0472, 0.0357, 0.0349], + device='cuda:0'), out_proj_covar=tensor([1.1157e-04, 9.1160e-05, 7.5166e-05, 1.0782e-04, 8.8158e-05, 1.4407e-04, + 1.0178e-04, 9.8844e-05], device='cuda:0') +2023-02-06 11:00:59,330 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.7201, 3.6491, 3.3361, 1.7594, 3.2820, 3.2937, 3.3686, 3.0881], + device='cuda:0'), covar=tensor([0.1080, 0.0796, 0.1156, 0.4876, 0.0958, 0.1200, 0.1442, 0.1132], + device='cuda:0'), in_proj_covar=tensor([0.0463, 0.0374, 0.0384, 0.0481, 0.0376, 0.0376, 0.0377, 0.0335], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 11:01:01,872 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=87724.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:01:06,417 INFO [train.py:901] (0/4) Epoch 11, batch 6900, loss[loss=0.1898, simple_loss=0.2647, pruned_loss=0.05746, over 7268.00 frames. ], tot_loss[loss=0.2362, simple_loss=0.3119, pruned_loss=0.08031, over 1617308.36 frames. ], batch size: 16, lr: 6.86e-03, grad_scale: 16.0 +2023-02-06 11:01:10,023 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87735.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 11:01:19,188 INFO [optim.py:369] (0/4) 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 11:01:26,848 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87760.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 11:01:41,613 INFO [train.py:901] (0/4) Epoch 11, batch 6950, loss[loss=0.2152, simple_loss=0.295, pruned_loss=0.06767, over 8364.00 frames. ], tot_loss[loss=0.2365, simple_loss=0.3119, pruned_loss=0.08059, over 1612129.41 frames. ], batch size: 24, lr: 6.85e-03, grad_scale: 16.0 +2023-02-06 11:01:52,585 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-06 11:02:11,619 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=87823.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:02:16,825 INFO [train.py:901] (0/4) Epoch 11, batch 7000, loss[loss=0.2106, simple_loss=0.2881, pruned_loss=0.06655, over 5134.00 frames. ], tot_loss[loss=0.2362, simple_loss=0.3112, pruned_loss=0.08054, over 1608602.86 frames. ], batch size: 11, lr: 6.85e-03, grad_scale: 16.0 +2023-02-06 11:02:22,311 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87839.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:02:25,937 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-02-06 11:02:29,501 INFO [optim.py:369] (0/4) 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 11:02:41,517 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 11, batch 7050, loss[loss=0.2457, simple_loss=0.3307, pruned_loss=0.08034, over 8497.00 frames. ], tot_loss[loss=0.2352, simple_loss=0.3104, pruned_loss=0.07995, over 1608890.13 frames. ], batch size: 49, lr: 6.85e-03, grad_scale: 16.0 +2023-02-06 11:03:26,708 INFO [train.py:901] (0/4) Epoch 11, batch 7100, loss[loss=0.2216, simple_loss=0.3092, pruned_loss=0.06704, over 8635.00 frames. ], tot_loss[loss=0.2337, simple_loss=0.3094, pruned_loss=0.07899, over 1612081.58 frames. ], batch size: 31, lr: 6.85e-03, grad_scale: 16.0 +2023-02-06 11:03:27,714 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-06 11:03:31,621 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87938.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:03:36,824 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 11:03:40,146 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=87951.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:03:48,274 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=87967.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:04:00,731 INFO [train.py:901] (0/4) Epoch 11, batch 7150, loss[loss=0.2647, simple_loss=0.3407, pruned_loss=0.09431, over 7643.00 frames. ], tot_loss[loss=0.2343, simple_loss=0.3098, pruned_loss=0.07935, over 1613502.52 frames. ], batch size: 19, lr: 6.85e-03, grad_scale: 16.0 +2023-02-06 11:04:01,614 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87982.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:04:05,829 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87988.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:04:14,502 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-88000.pt +2023-02-06 11:04:36,670 INFO [train.py:901] (0/4) Epoch 11, batch 7200, loss[loss=0.278, simple_loss=0.3491, pruned_loss=0.1034, over 8515.00 frames. ], tot_loss[loss=0.2352, simple_loss=0.3102, pruned_loss=0.08011, over 1614715.44 frames. ], batch size: 28, lr: 6.84e-03, grad_scale: 32.0 +2023-02-06 11:04:49,439 INFO [optim.py:369] (0/4) 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 11:04:57,778 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88061.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:05:01,275 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88066.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:05:10,857 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 +2023-02-06 11:05:11,823 INFO [train.py:901] (0/4) Epoch 11, batch 7250, loss[loss=0.2288, simple_loss=0.3006, pruned_loss=0.07848, over 8561.00 frames. ], tot_loss[loss=0.2354, simple_loss=0.3109, pruned_loss=0.07996, over 1618461.56 frames. ], batch size: 39, lr: 6.84e-03, grad_scale: 32.0 +2023-02-06 11:05:12,649 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88082.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:05:19,795 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 +2023-02-06 11:05:21,492 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=88095.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:05:37,924 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=88118.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:05:39,356 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=88120.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:05:46,928 INFO [train.py:901] (0/4) Epoch 11, batch 7300, loss[loss=0.2698, simple_loss=0.3281, pruned_loss=0.1057, over 7785.00 frames. ], tot_loss[loss=0.2361, simple_loss=0.3111, pruned_loss=0.08052, over 1614166.69 frames. ], batch size: 19, lr: 6.84e-03, grad_scale: 32.0 +2023-02-06 11:06:00,702 INFO [optim.py:369] (0/4) 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 11:06:22,828 INFO [train.py:901] (0/4) Epoch 11, batch 7350, loss[loss=0.246, simple_loss=0.3144, pruned_loss=0.08877, over 8405.00 frames. ], tot_loss[loss=0.2364, simple_loss=0.3115, pruned_loss=0.08066, over 1613738.92 frames. ], batch size: 49, lr: 6.84e-03, grad_scale: 32.0 +2023-02-06 11:06:28,427 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8048, 1.4266, 3.4175, 1.4601, 2.1816, 3.8031, 3.8330, 3.2511], + device='cuda:0'), covar=tensor([0.1262, 0.1734, 0.0407, 0.2116, 0.1224, 0.0230, 0.0476, 0.0647], + device='cuda:0'), in_proj_covar=tensor([0.0262, 0.0296, 0.0258, 0.0291, 0.0273, 0.0233, 0.0336, 0.0288], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-06 11:06:32,278 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-06 11:06:32,480 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=88194.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:06:50,338 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=88219.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:06:51,489 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 11, batch 7400, loss[loss=0.2581, simple_loss=0.3412, pruned_loss=0.08748, over 8623.00 frames. ], tot_loss[loss=0.2367, simple_loss=0.312, pruned_loss=0.08069, over 1615489.94 frames. ], batch size: 34, lr: 6.84e-03, grad_scale: 16.0 +2023-02-06 11:07:03,172 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=88238.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:07:11,886 INFO [optim.py:369] (0/4) 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 11:07:20,893 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=88263.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:07:32,898 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 11, batch 7450, loss[loss=0.2343, simple_loss=0.3121, pruned_loss=0.07828, over 8586.00 frames. ], tot_loss[loss=0.2355, simple_loss=0.3111, pruned_loss=0.07994, over 1617606.07 frames. ], batch size: 34, lr: 6.83e-03, grad_scale: 16.0 +2023-02-06 11:07:43,215 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4464, 1.9101, 3.3660, 1.2198, 2.3901, 1.9035, 1.5216, 2.4157], + device='cuda:0'), covar=tensor([0.1794, 0.2362, 0.0697, 0.4139, 0.1626, 0.2962, 0.1968, 0.2161], + device='cuda:0'), in_proj_covar=tensor([0.0489, 0.0522, 0.0530, 0.0578, 0.0620, 0.0559, 0.0473, 0.0620], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 11:07:52,582 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=88309.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 11:07:58,566 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=88317.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:08:01,841 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=88322.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:08:07,731 INFO [train.py:901] (0/4) Epoch 11, batch 7500, loss[loss=0.2648, simple_loss=0.3242, pruned_loss=0.1027, over 6602.00 frames. ], tot_loss[loss=0.2344, simple_loss=0.3101, pruned_loss=0.07938, over 1612645.38 frames. ], batch size: 71, lr: 6.83e-03, grad_scale: 16.0 +2023-02-06 11:08:13,321 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=88338.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:08:15,868 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=88342.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:08:19,204 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=88347.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:08:20,934 INFO [optim.py:369] (0/4) 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 11:08:29,978 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=88363.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:08:42,585 INFO [train.py:901] (0/4) Epoch 11, batch 7550, loss[loss=0.244, simple_loss=0.3205, pruned_loss=0.08378, over 8506.00 frames. ], tot_loss[loss=0.2348, simple_loss=0.3103, pruned_loss=0.07962, over 1616113.90 frames. ], batch size: 26, lr: 6.83e-03, grad_scale: 16.0 +2023-02-06 11:09:12,948 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8582, 1.4291, 1.6317, 1.3429, 1.0748, 1.4182, 1.6340, 1.3135], + device='cuda:0'), covar=tensor([0.0515, 0.1240, 0.1666, 0.1409, 0.0573, 0.1498, 0.0691, 0.0650], + device='cuda:0'), in_proj_covar=tensor([0.0100, 0.0150, 0.0190, 0.0157, 0.0103, 0.0163, 0.0115, 0.0137], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0006], + device='cuda:0') +2023-02-06 11:09:17,375 INFO [train.py:901] (0/4) Epoch 11, batch 7600, loss[loss=0.2603, simple_loss=0.328, pruned_loss=0.09626, over 8569.00 frames. ], tot_loss[loss=0.2338, simple_loss=0.3096, pruned_loss=0.07899, over 1612983.00 frames. ], batch size: 34, lr: 6.83e-03, grad_scale: 16.0 +2023-02-06 11:09:29,096 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.4754, 1.4800, 4.7044, 1.7945, 4.1080, 3.9355, 4.3296, 4.1211], + device='cuda:0'), covar=tensor([0.0545, 0.4463, 0.0445, 0.3293, 0.1069, 0.0863, 0.0444, 0.0612], + device='cuda:0'), in_proj_covar=tensor([0.0496, 0.0582, 0.0595, 0.0539, 0.0615, 0.0524, 0.0515, 0.0587], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 11:09:31,031 INFO [optim.py:369] (0/4) 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] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=88462.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:09:51,502 INFO [train.py:901] (0/4) Epoch 11, batch 7650, loss[loss=0.2101, simple_loss=0.273, pruned_loss=0.07366, over 7791.00 frames. ], tot_loss[loss=0.2335, simple_loss=0.309, pruned_loss=0.07904, over 1611157.79 frames. ], batch size: 19, lr: 6.83e-03, grad_scale: 16.0 +2023-02-06 11:10:23,423 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.48 vs. limit=5.0 +2023-02-06 11:10:26,448 INFO [train.py:901] (0/4) Epoch 11, batch 7700, loss[loss=0.2202, simple_loss=0.2965, pruned_loss=0.07194, over 8360.00 frames. ], tot_loss[loss=0.232, simple_loss=0.3076, pruned_loss=0.07824, over 1611318.39 frames. ], batch size: 24, lr: 6.82e-03, grad_scale: 16.0 +2023-02-06 11:10:39,166 WARNING [train.py:1067] (0/4) 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] (0/4) 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 11:10:58,417 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 11, batch 7750, loss[loss=0.2015, simple_loss=0.2871, pruned_loss=0.05795, over 8032.00 frames. ], tot_loss[loss=0.2343, simple_loss=0.3092, pruned_loss=0.07968, over 1613370.34 frames. ], batch size: 22, lr: 6.82e-03, grad_scale: 16.0 +2023-02-06 11:11:36,341 INFO [train.py:901] (0/4) Epoch 11, batch 7800, loss[loss=0.2479, simple_loss=0.3287, pruned_loss=0.08355, over 8012.00 frames. ], tot_loss[loss=0.2353, simple_loss=0.31, pruned_loss=0.08026, over 1611785.69 frames. ], batch size: 22, lr: 6.82e-03, grad_scale: 16.0 +2023-02-06 11:11:48,834 INFO [optim.py:369] (0/4) 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] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=88653.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 11:12:04,960 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2085, 1.1062, 1.2729, 1.1473, 1.0012, 1.3232, 0.0848, 0.9930], + device='cuda:0'), covar=tensor([0.2408, 0.1849, 0.0677, 0.1340, 0.3784, 0.0615, 0.3264, 0.1691], + device='cuda:0'), in_proj_covar=tensor([0.0164, 0.0169, 0.0101, 0.0212, 0.0253, 0.0103, 0.0164, 0.0165], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 11:12:09,483 INFO [train.py:901] (0/4) Epoch 11, batch 7850, loss[loss=0.2822, simple_loss=0.3565, pruned_loss=0.1039, over 8354.00 frames. ], tot_loss[loss=0.2355, simple_loss=0.3105, pruned_loss=0.08026, over 1611891.26 frames. ], batch size: 24, lr: 6.82e-03, grad_scale: 16.0 +2023-02-06 11:12:42,904 INFO [train.py:901] (0/4) Epoch 11, batch 7900, loss[loss=0.2371, simple_loss=0.3193, pruned_loss=0.07747, over 8554.00 frames. ], tot_loss[loss=0.2334, simple_loss=0.3088, pruned_loss=0.079, over 1610235.92 frames. ], batch size: 31, lr: 6.82e-03, grad_scale: 16.0 +2023-02-06 11:12:55,426 INFO [optim.py:369] (0/4) 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 11:13:04,849 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.66 vs. limit=5.0 +2023-02-06 11:13:07,248 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88768.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 11:13:15,828 INFO [train.py:901] (0/4) Epoch 11, batch 7950, loss[loss=0.2454, simple_loss=0.3244, pruned_loss=0.08322, over 8488.00 frames. ], tot_loss[loss=0.2329, simple_loss=0.3083, pruned_loss=0.0787, over 1609559.48 frames. ], batch size: 27, lr: 6.81e-03, grad_scale: 16.0 +2023-02-06 11:13:24,834 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7116, 1.5776, 1.7773, 1.3628, 0.9242, 1.4896, 1.6575, 1.7292], + device='cuda:0'), covar=tensor([0.0528, 0.1119, 0.1574, 0.1320, 0.0570, 0.1457, 0.0614, 0.0508], + device='cuda:0'), in_proj_covar=tensor([0.0100, 0.0150, 0.0191, 0.0157, 0.0103, 0.0164, 0.0115, 0.0137], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0006], + device='cuda:0') +2023-02-06 11:13:49,362 INFO [train.py:901] (0/4) Epoch 11, batch 8000, loss[loss=0.207, simple_loss=0.2822, pruned_loss=0.06589, over 7695.00 frames. ], tot_loss[loss=0.2333, simple_loss=0.3082, pruned_loss=0.07915, over 1605654.32 frames. ], batch size: 18, lr: 6.81e-03, grad_scale: 16.0 +2023-02-06 11:13:50,911 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=88833.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:13:54,535 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 +2023-02-06 11:14:02,011 INFO [optim.py:369] (0/4) 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 11:14:07,223 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=88858.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:14:22,230 INFO [train.py:901] (0/4) Epoch 11, batch 8050, loss[loss=0.1773, simple_loss=0.26, pruned_loss=0.04733, over 7527.00 frames. ], tot_loss[loss=0.2329, simple_loss=0.3072, pruned_loss=0.07932, over 1586978.86 frames. ], batch size: 18, lr: 6.81e-03, grad_scale: 16.0 +2023-02-06 11:14:44,457 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/epoch-11.pt +2023-02-06 11:14:55,381 WARNING [train.py:1067] (0/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-06 11:14:58,684 INFO [train.py:901] (0/4) Epoch 12, batch 0, loss[loss=0.2693, simple_loss=0.3484, pruned_loss=0.09507, over 8108.00 frames. ], tot_loss[loss=0.2693, simple_loss=0.3484, pruned_loss=0.09507, over 8108.00 frames. ], batch size: 23, lr: 6.52e-03, grad_scale: 16.0 +2023-02-06 11:14:58,685 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 11:15:09,781 INFO [train.py:935] (0/4) Epoch 12, validation: loss=0.1897, simple_loss=0.2896, pruned_loss=0.04486, over 944034.00 frames. +2023-02-06 11:15:09,782 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-06 11:15:23,298 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-06 11:15:28,755 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4991, 2.7036, 1.8306, 2.0905, 2.1446, 1.5389, 2.0098, 2.0332], + device='cuda:0'), covar=tensor([0.1273, 0.0282, 0.1007, 0.0590, 0.0573, 0.1220, 0.0815, 0.0800], + device='cuda:0'), in_proj_covar=tensor([0.0348, 0.0236, 0.0318, 0.0299, 0.0303, 0.0323, 0.0341, 0.0307], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 11:15:35,201 INFO [optim.py:369] (0/4) 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 11:15:44,685 INFO [train.py:901] (0/4) Epoch 12, batch 50, loss[loss=0.2391, simple_loss=0.3207, pruned_loss=0.07874, over 8502.00 frames. ], tot_loss[loss=0.2486, simple_loss=0.3214, pruned_loss=0.08787, over 367002.54 frames. ], batch size: 26, lr: 6.52e-03, grad_scale: 16.0 +2023-02-06 11:15:57,432 WARNING [train.py:1067] (0/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-06 11:16:19,036 WARNING [train.py:1067] (0/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-06 11:16:19,746 INFO [train.py:901] (0/4) Epoch 12, batch 100, loss[loss=0.2716, simple_loss=0.3468, pruned_loss=0.09818, over 8339.00 frames. ], tot_loss[loss=0.2458, simple_loss=0.319, pruned_loss=0.08626, over 646886.39 frames. ], batch size: 26, lr: 6.52e-03, grad_scale: 16.0 +2023-02-06 11:16:26,510 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89024.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 11:16:40,635 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89045.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:16:43,450 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89049.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 11:16:43,912 INFO [optim.py:369] (0/4) 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 11:16:54,736 INFO [train.py:901] (0/4) Epoch 12, batch 150, loss[loss=0.2282, simple_loss=0.3122, pruned_loss=0.07213, over 8031.00 frames. ], tot_loss[loss=0.2427, simple_loss=0.3162, pruned_loss=0.08456, over 860570.98 frames. ], batch size: 22, lr: 6.52e-03, grad_scale: 16.0 +2023-02-06 11:17:00,755 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.7293, 4.6558, 4.2406, 1.9555, 4.2037, 4.2540, 4.2937, 4.1164], + device='cuda:0'), covar=tensor([0.0732, 0.0570, 0.0944, 0.4785, 0.0752, 0.0852, 0.1248, 0.0769], + device='cuda:0'), in_proj_covar=tensor([0.0462, 0.0370, 0.0382, 0.0476, 0.0372, 0.0375, 0.0373, 0.0332], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 11:17:29,010 INFO [train.py:901] (0/4) Epoch 12, batch 200, loss[loss=0.2108, simple_loss=0.2697, pruned_loss=0.07594, over 7536.00 frames. ], tot_loss[loss=0.2388, simple_loss=0.3122, pruned_loss=0.08269, over 1022664.26 frames. ], batch size: 18, lr: 6.52e-03, grad_scale: 16.0 +2023-02-06 11:17:33,288 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1079, 1.7200, 2.4374, 2.0686, 2.1914, 1.9956, 1.6386, 1.2091], + device='cuda:0'), covar=tensor([0.3421, 0.3678, 0.1119, 0.2160, 0.1726, 0.2044, 0.1516, 0.3332], + device='cuda:0'), in_proj_covar=tensor([0.0873, 0.0858, 0.0721, 0.0830, 0.0916, 0.0789, 0.0692, 0.0750], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 11:17:38,637 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1027, 1.5208, 1.5059, 1.1696, 0.9469, 1.3755, 1.6438, 1.6675], + device='cuda:0'), covar=tensor([0.0508, 0.1247, 0.1811, 0.1478, 0.0611, 0.1568, 0.0707, 0.0569], + device='cuda:0'), in_proj_covar=tensor([0.0100, 0.0152, 0.0191, 0.0157, 0.0103, 0.0163, 0.0115, 0.0137], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0006], + device='cuda:0') +2023-02-06 11:17:53,942 INFO [optim.py:369] (0/4) 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 11:18:03,556 INFO [train.py:901] (0/4) Epoch 12, batch 250, loss[loss=0.2501, simple_loss=0.3281, pruned_loss=0.08608, over 8464.00 frames. ], tot_loss[loss=0.237, simple_loss=0.3114, pruned_loss=0.08129, over 1152148.06 frames. ], batch size: 25, lr: 6.51e-03, grad_scale: 16.0 +2023-02-06 11:18:04,453 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8534, 2.0628, 1.7349, 2.5913, 1.0481, 1.4245, 1.6137, 2.2153], + device='cuda:0'), covar=tensor([0.0804, 0.0843, 0.1052, 0.0401, 0.1222, 0.1531, 0.1048, 0.0728], + device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0216, 0.0254, 0.0214, 0.0218, 0.0252, 0.0258, 0.0222], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 11:18:13,252 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-06 11:18:20,896 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89187.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:18:22,859 WARNING [train.py:1067] (0/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-06 11:18:38,842 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9104, 1.5936, 1.6288, 1.3657, 0.9853, 1.4862, 1.6883, 1.3743], + device='cuda:0'), covar=tensor([0.0516, 0.1149, 0.1628, 0.1353, 0.0629, 0.1438, 0.0671, 0.0627], + device='cuda:0'), in_proj_covar=tensor([0.0100, 0.0152, 0.0192, 0.0158, 0.0103, 0.0163, 0.0115, 0.0137], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0006], + device='cuda:0') +2023-02-06 11:18:40,049 INFO [train.py:901] (0/4) Epoch 12, batch 300, loss[loss=0.1803, simple_loss=0.2608, pruned_loss=0.04993, over 8292.00 frames. ], tot_loss[loss=0.2368, simple_loss=0.3117, pruned_loss=0.08091, over 1254391.03 frames. ], batch size: 23, lr: 6.51e-03, grad_scale: 16.0 +2023-02-06 11:19:04,996 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 12, batch 350, loss[loss=0.2002, simple_loss=0.271, pruned_loss=0.06467, over 7517.00 frames. ], tot_loss[loss=0.2374, simple_loss=0.3123, pruned_loss=0.08118, over 1333128.77 frames. ], batch size: 18, lr: 6.51e-03, grad_scale: 16.0 +2023-02-06 11:19:32,083 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0770, 1.2594, 1.2236, 0.6306, 1.2241, 1.0241, 0.0614, 1.1322], + device='cuda:0'), covar=tensor([0.0231, 0.0205, 0.0185, 0.0304, 0.0220, 0.0532, 0.0488, 0.0177], + device='cuda:0'), in_proj_covar=tensor([0.0389, 0.0326, 0.0273, 0.0381, 0.0314, 0.0472, 0.0356, 0.0352], + device='cuda:0'), out_proj_covar=tensor([1.1072e-04, 9.0777e-05, 7.6165e-05, 1.0662e-04, 8.8719e-05, 1.4344e-04, + 1.0119e-04, 9.9422e-05], device='cuda:0') +2023-02-06 11:19:39,013 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.06 vs. limit=5.0 +2023-02-06 11:19:49,366 INFO [train.py:901] (0/4) Epoch 12, batch 400, loss[loss=0.2101, simple_loss=0.2925, pruned_loss=0.06383, over 8025.00 frames. ], tot_loss[loss=0.2364, simple_loss=0.3115, pruned_loss=0.08066, over 1393869.62 frames. ], batch size: 22, lr: 6.51e-03, grad_scale: 16.0 +2023-02-06 11:20:14,267 INFO [optim.py:369] (0/4) 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 11:20:19,706 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([6.0670, 1.5838, 6.1643, 2.2676, 5.6508, 5.2231, 5.7646, 5.5517], + device='cuda:0'), covar=tensor([0.0436, 0.4125, 0.0238, 0.3001, 0.0736, 0.0700, 0.0359, 0.0413], + device='cuda:0'), in_proj_covar=tensor([0.0490, 0.0571, 0.0579, 0.0529, 0.0606, 0.0512, 0.0507, 0.0576], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 11:20:24,226 INFO [train.py:901] (0/4) Epoch 12, batch 450, loss[loss=0.2554, simple_loss=0.3332, pruned_loss=0.08884, over 8571.00 frames. ], tot_loss[loss=0.2356, simple_loss=0.3112, pruned_loss=0.07996, over 1445690.82 frames. ], batch size: 31, lr: 6.51e-03, grad_scale: 16.0 +2023-02-06 11:20:40,992 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=89389.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:20:43,193 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89392.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:20:58,875 INFO [train.py:901] (0/4) Epoch 12, batch 500, loss[loss=0.1937, simple_loss=0.2782, pruned_loss=0.05465, over 7928.00 frames. ], tot_loss[loss=0.234, simple_loss=0.3095, pruned_loss=0.07931, over 1482587.55 frames. ], batch size: 20, lr: 6.51e-03, grad_scale: 16.0 +2023-02-06 11:21:00,672 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 +2023-02-06 11:21:15,394 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9940, 1.5160, 3.4798, 1.5542, 2.4440, 3.8612, 4.0138, 3.3800], + device='cuda:0'), covar=tensor([0.1027, 0.1494, 0.0347, 0.1804, 0.0926, 0.0246, 0.0382, 0.0574], + device='cuda:0'), in_proj_covar=tensor([0.0266, 0.0296, 0.0261, 0.0293, 0.0273, 0.0237, 0.0343, 0.0291], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 11:21:19,402 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89443.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:21:22,152 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0774, 1.6037, 4.3538, 1.9488, 2.4063, 4.9024, 4.9692, 4.3167], + device='cuda:0'), covar=tensor([0.1134, 0.1552, 0.0264, 0.1767, 0.1048, 0.0206, 0.0353, 0.0552], + device='cuda:0'), in_proj_covar=tensor([0.0266, 0.0296, 0.0261, 0.0293, 0.0273, 0.0237, 0.0343, 0.0291], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 11:21:24,108 INFO [optim.py:369] (0/4) 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 11:21:29,683 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89457.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 11:21:34,349 INFO [train.py:901] (0/4) Epoch 12, batch 550, loss[loss=0.2128, simple_loss=0.2869, pruned_loss=0.06937, over 7640.00 frames. ], tot_loss[loss=0.2329, simple_loss=0.3084, pruned_loss=0.07866, over 1511581.68 frames. ], batch size: 19, lr: 6.50e-03, grad_scale: 16.0 +2023-02-06 11:22:02,579 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 12, batch 600, loss[loss=0.1959, simple_loss=0.2781, pruned_loss=0.0569, over 7810.00 frames. ], tot_loss[loss=0.2335, simple_loss=0.3094, pruned_loss=0.07879, over 1538717.46 frames. ], batch size: 19, lr: 6.50e-03, grad_scale: 16.0 +2023-02-06 11:22:17,842 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89527.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:22:20,451 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=89531.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:22:26,551 WARNING [train.py:1067] (0/4) 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] (0/4) 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 11:22:41,118 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.70 vs. limit=5.0 +2023-02-06 11:22:44,045 INFO [train.py:901] (0/4) Epoch 12, batch 650, loss[loss=0.2311, simple_loss=0.3117, pruned_loss=0.07528, over 8468.00 frames. ], tot_loss[loss=0.2332, simple_loss=0.3096, pruned_loss=0.07836, over 1559230.54 frames. ], batch size: 29, lr: 6.50e-03, grad_scale: 16.0 +2023-02-06 11:22:54,319 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4272, 2.2245, 1.7808, 1.9634, 1.8503, 1.4340, 1.7208, 1.8286], + device='cuda:0'), covar=tensor([0.0879, 0.0294, 0.0833, 0.0447, 0.0514, 0.1147, 0.0682, 0.0549], + device='cuda:0'), in_proj_covar=tensor([0.0343, 0.0237, 0.0316, 0.0296, 0.0300, 0.0321, 0.0336, 0.0304], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 11:23:18,861 INFO [train.py:901] (0/4) Epoch 12, batch 700, loss[loss=0.2476, simple_loss=0.3245, pruned_loss=0.08537, over 8514.00 frames. ], tot_loss[loss=0.2335, simple_loss=0.3096, pruned_loss=0.07868, over 1574401.74 frames. ], batch size: 26, lr: 6.50e-03, grad_scale: 16.0 +2023-02-06 11:23:22,434 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5188, 1.4350, 2.7203, 1.2078, 2.0383, 2.9012, 3.0981, 2.3319], + device='cuda:0'), covar=tensor([0.1204, 0.1560, 0.0467, 0.2182, 0.0956, 0.0395, 0.0629, 0.0859], + device='cuda:0'), in_proj_covar=tensor([0.0268, 0.0298, 0.0263, 0.0294, 0.0275, 0.0238, 0.0345, 0.0293], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 11:23:39,701 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.4994, 1.4139, 4.6985, 1.6122, 4.0907, 3.8969, 4.1975, 4.0165], + device='cuda:0'), covar=tensor([0.0539, 0.4389, 0.0424, 0.3480, 0.1104, 0.0860, 0.0574, 0.0659], + device='cuda:0'), in_proj_covar=tensor([0.0495, 0.0575, 0.0582, 0.0530, 0.0607, 0.0517, 0.0510, 0.0579], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 11:23:40,391 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=89646.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:23:43,622 INFO [optim.py:369] (0/4) 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 11:23:53,828 INFO [train.py:901] (0/4) Epoch 12, batch 750, loss[loss=0.28, simple_loss=0.3513, pruned_loss=0.1044, over 8327.00 frames. ], tot_loss[loss=0.2336, simple_loss=0.3096, pruned_loss=0.07884, over 1585502.85 frames. ], batch size: 25, lr: 6.50e-03, grad_scale: 16.0 +2023-02-06 11:23:54,685 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8084, 2.3098, 1.7701, 2.7461, 1.3253, 1.5777, 1.7695, 2.3024], + device='cuda:0'), covar=tensor([0.0841, 0.0727, 0.1006, 0.0374, 0.1140, 0.1402, 0.1065, 0.0735], + device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0212, 0.0254, 0.0213, 0.0217, 0.0251, 0.0257, 0.0219], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 11:23:59,424 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0289, 3.9779, 2.5613, 2.5458, 2.7663, 2.0221, 2.6181, 2.8733], + device='cuda:0'), covar=tensor([0.1498, 0.0246, 0.0802, 0.0771, 0.0648, 0.1179, 0.1122, 0.1041], + device='cuda:0'), in_proj_covar=tensor([0.0349, 0.0241, 0.0319, 0.0300, 0.0303, 0.0325, 0.0340, 0.0308], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 11:24:11,501 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-06 11:24:17,636 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89698.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:24:20,262 WARNING [train.py:1067] (0/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-06 11:24:23,777 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.7193, 2.3012, 3.4068, 2.5442, 2.9227, 2.5574, 2.1040, 1.8544], + device='cuda:0'), covar=tensor([0.3437, 0.3980, 0.1098, 0.2591, 0.1983, 0.2047, 0.1643, 0.3837], + device='cuda:0'), in_proj_covar=tensor([0.0884, 0.0865, 0.0733, 0.0840, 0.0925, 0.0795, 0.0700, 0.0764], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 11:24:28,197 INFO [train.py:901] (0/4) Epoch 12, batch 800, loss[loss=0.2243, simple_loss=0.3178, pruned_loss=0.06539, over 8293.00 frames. ], tot_loss[loss=0.2335, simple_loss=0.3093, pruned_loss=0.07884, over 1594602.75 frames. ], batch size: 23, lr: 6.49e-03, grad_scale: 8.0 +2023-02-06 11:24:32,957 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5667, 1.5843, 2.9134, 1.1921, 2.1047, 3.1937, 3.3960, 2.3374], + device='cuda:0'), covar=tensor([0.1485, 0.1641, 0.0445, 0.2592, 0.1037, 0.0439, 0.0533, 0.1140], + device='cuda:0'), in_proj_covar=tensor([0.0269, 0.0300, 0.0263, 0.0296, 0.0275, 0.0239, 0.0347, 0.0292], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 11:24:33,180 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 +2023-02-06 11:24:43,641 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=89736.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:24:53,390 INFO [optim.py:369] (0/4) 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 11:24:59,645 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 12, batch 850, loss[loss=0.225, simple_loss=0.3018, pruned_loss=0.07413, over 8137.00 frames. ], tot_loss[loss=0.2339, simple_loss=0.3098, pruned_loss=0.07898, over 1602745.83 frames. ], batch size: 22, lr: 6.49e-03, grad_scale: 8.0 +2023-02-06 11:25:17,992 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89785.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:25:19,267 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=89801.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 11:25:32,942 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.54 vs. limit=5.0 +2023-02-06 11:25:37,800 INFO [train.py:901] (0/4) Epoch 12, batch 900, loss[loss=0.2434, simple_loss=0.3007, pruned_loss=0.09306, over 7234.00 frames. ], tot_loss[loss=0.2335, simple_loss=0.3093, pruned_loss=0.0788, over 1605532.31 frames. ], batch size: 16, lr: 6.49e-03, grad_scale: 8.0 +2023-02-06 11:26:03,293 INFO [optim.py:369] (0/4) 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 11:26:03,485 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=89851.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:26:11,297 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6677, 1.8751, 1.5930, 2.2818, 0.9894, 1.4091, 1.5291, 1.8515], + device='cuda:0'), covar=tensor([0.0852, 0.0847, 0.1025, 0.0471, 0.1214, 0.1516, 0.1100, 0.0822], + device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0213, 0.0255, 0.0215, 0.0218, 0.0252, 0.0258, 0.0218], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 11:26:11,800 INFO [train.py:901] (0/4) Epoch 12, batch 950, loss[loss=0.1906, simple_loss=0.2725, pruned_loss=0.0544, over 8085.00 frames. ], tot_loss[loss=0.2331, simple_loss=0.3094, pruned_loss=0.07846, over 1609960.11 frames. ], batch size: 21, lr: 6.49e-03, grad_scale: 8.0 +2023-02-06 11:26:16,405 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=89871.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:26:24,524 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89883.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:26:33,234 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3999, 1.8972, 2.9945, 2.2080, 2.6754, 2.1770, 1.7484, 1.2772], + device='cuda:0'), covar=tensor([0.4296, 0.4452, 0.1200, 0.2866, 0.1938, 0.2381, 0.1797, 0.4547], + device='cuda:0'), in_proj_covar=tensor([0.0895, 0.0876, 0.0743, 0.0853, 0.0936, 0.0806, 0.0708, 0.0772], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 11:26:38,568 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=89902.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:26:38,609 INFO [zipformer.py:1185] (0/4) 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] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-06 11:26:46,281 INFO [train.py:901] (0/4) Epoch 12, batch 1000, loss[loss=0.2668, simple_loss=0.3362, pruned_loss=0.09871, over 8370.00 frames. ], tot_loss[loss=0.2325, simple_loss=0.3088, pruned_loss=0.07807, over 1612999.79 frames. ], batch size: 24, lr: 6.49e-03, grad_scale: 8.0 +2023-02-06 11:26:47,819 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=89916.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 11:26:48,419 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6507, 1.6057, 2.2035, 1.7354, 1.1782, 2.1500, 0.3599, 1.3861], + device='cuda:0'), covar=tensor([0.2576, 0.1887, 0.0472, 0.1557, 0.4472, 0.0469, 0.3385, 0.1837], + device='cuda:0'), in_proj_covar=tensor([0.0170, 0.0172, 0.0103, 0.0218, 0.0259, 0.0107, 0.0166, 0.0169], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 11:26:55,681 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89927.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:27:11,392 INFO [optim.py:369] (0/4) 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 11:27:11,418 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-06 11:27:13,883 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 +2023-02-06 11:27:20,836 INFO [train.py:901] (0/4) Epoch 12, batch 1050, loss[loss=0.208, simple_loss=0.2977, pruned_loss=0.05912, over 8137.00 frames. ], tot_loss[loss=0.2332, simple_loss=0.3089, pruned_loss=0.07873, over 1613583.16 frames. ], batch size: 22, lr: 6.49e-03, grad_scale: 8.0 +2023-02-06 11:27:22,337 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.0736, 1.1625, 4.3372, 1.5639, 3.6944, 3.6398, 3.9082, 3.7269], + device='cuda:0'), covar=tensor([0.0595, 0.4677, 0.0488, 0.3464, 0.1222, 0.0838, 0.0542, 0.0638], + device='cuda:0'), in_proj_covar=tensor([0.0501, 0.0586, 0.0593, 0.0543, 0.0620, 0.0528, 0.0520, 0.0592], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 11:27:24,323 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-06 11:27:35,822 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=89986.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:27:44,987 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-90000.pt +2023-02-06 11:27:56,226 INFO [train.py:901] (0/4) Epoch 12, batch 1100, loss[loss=0.2659, simple_loss=0.3288, pruned_loss=0.1015, over 8353.00 frames. ], tot_loss[loss=0.2341, simple_loss=0.3095, pruned_loss=0.07936, over 1613019.04 frames. ], batch size: 49, lr: 6.48e-03, grad_scale: 8.0 +2023-02-06 11:28:04,623 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2782, 2.4573, 1.7165, 2.0116, 1.9191, 1.3990, 1.7966, 1.8427], + device='cuda:0'), covar=tensor([0.1460, 0.0389, 0.1129, 0.0608, 0.0699, 0.1431, 0.0940, 0.0877], + device='cuda:0'), in_proj_covar=tensor([0.0345, 0.0240, 0.0316, 0.0297, 0.0301, 0.0323, 0.0336, 0.0306], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 11:28:15,783 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 11:28:31,017 INFO [train.py:901] (0/4) Epoch 12, batch 1150, loss[loss=0.281, simple_loss=0.3461, pruned_loss=0.108, over 8188.00 frames. ], tot_loss[loss=0.234, simple_loss=0.3093, pruned_loss=0.07936, over 1616107.91 frames. ], batch size: 23, lr: 6.48e-03, grad_scale: 4.0 +2023-02-06 11:28:34,436 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-06 11:28:41,704 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-02-06 11:28:59,641 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.6965, 1.2177, 4.8303, 1.7118, 4.3429, 4.0638, 4.4247, 4.2363], + device='cuda:0'), covar=tensor([0.0443, 0.4553, 0.0449, 0.3508, 0.0939, 0.0783, 0.0456, 0.0561], + device='cuda:0'), in_proj_covar=tensor([0.0494, 0.0581, 0.0589, 0.0534, 0.0611, 0.0521, 0.0512, 0.0584], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 11:29:01,091 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90107.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:29:05,603 INFO [train.py:901] (0/4) Epoch 12, batch 1200, loss[loss=0.2413, simple_loss=0.3226, pruned_loss=0.08002, over 8610.00 frames. ], tot_loss[loss=0.2339, simple_loss=0.3096, pruned_loss=0.07909, over 1618355.39 frames. ], batch size: 34, lr: 6.48e-03, grad_scale: 8.0 +2023-02-06 11:29:19,572 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90132.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:29:30,356 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 11:29:36,546 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90157.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:29:37,310 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90158.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:29:41,827 INFO [train.py:901] (0/4) Epoch 12, batch 1250, loss[loss=0.2337, simple_loss=0.2973, pruned_loss=0.08509, over 7830.00 frames. ], tot_loss[loss=0.2336, simple_loss=0.3092, pruned_loss=0.07904, over 1615881.46 frames. ], batch size: 18, lr: 6.48e-03, grad_scale: 8.0 +2023-02-06 11:29:47,513 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90172.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 11:29:55,711 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90183.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:30:05,433 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90197.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 11:30:17,132 INFO [train.py:901] (0/4) Epoch 12, batch 1300, loss[loss=0.231, simple_loss=0.2919, pruned_loss=0.08503, over 7663.00 frames. ], tot_loss[loss=0.2335, simple_loss=0.3093, pruned_loss=0.07888, over 1616068.16 frames. ], batch size: 19, lr: 6.48e-03, grad_scale: 8.0 +2023-02-06 11:30:26,160 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=90227.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:30:37,500 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90242.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:30:37,677 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.63 vs. limit=2.0 +2023-02-06 11:30:43,674 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90250.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:30:44,949 INFO [optim.py:369] (0/4) 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 11:30:53,239 INFO [train.py:901] (0/4) Epoch 12, batch 1350, loss[loss=0.2117, simple_loss=0.2743, pruned_loss=0.07452, over 7248.00 frames. ], tot_loss[loss=0.2338, simple_loss=0.3099, pruned_loss=0.07886, over 1619719.78 frames. ], batch size: 16, lr: 6.47e-03, grad_scale: 8.0 +2023-02-06 11:30:55,575 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90267.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:31:28,606 INFO [train.py:901] (0/4) Epoch 12, batch 1400, loss[loss=0.2373, simple_loss=0.3014, pruned_loss=0.08655, over 8280.00 frames. ], tot_loss[loss=0.2332, simple_loss=0.3092, pruned_loss=0.07858, over 1622148.05 frames. ], batch size: 23, lr: 6.47e-03, grad_scale: 8.0 +2023-02-06 11:31:47,888 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 11:32:03,610 INFO [train.py:901] (0/4) Epoch 12, batch 1450, loss[loss=0.2481, simple_loss=0.3015, pruned_loss=0.09737, over 7928.00 frames. ], tot_loss[loss=0.2326, simple_loss=0.3085, pruned_loss=0.07831, over 1621149.35 frames. ], batch size: 20, lr: 6.47e-03, grad_scale: 8.0 +2023-02-06 11:32:07,900 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90369.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:32:08,439 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-06 11:32:25,350 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3534, 1.8453, 1.9169, 1.7751, 1.3434, 1.8433, 2.0919, 2.0034], + device='cuda:0'), covar=tensor([0.0505, 0.0971, 0.1412, 0.1172, 0.0591, 0.1204, 0.0627, 0.0474], + device='cuda:0'), in_proj_covar=tensor([0.0100, 0.0152, 0.0191, 0.0158, 0.0103, 0.0163, 0.0116, 0.0137], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0006], + device='cuda:0') +2023-02-06 11:32:38,140 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90413.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:32:38,602 INFO [train.py:901] (0/4) Epoch 12, batch 1500, loss[loss=0.208, simple_loss=0.2925, pruned_loss=0.06171, over 8187.00 frames. ], tot_loss[loss=0.2332, simple_loss=0.309, pruned_loss=0.07864, over 1621369.64 frames. ], batch size: 23, lr: 6.47e-03, grad_scale: 8.0 +2023-02-06 11:32:55,126 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90438.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:33:04,321 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 12, batch 1550, loss[loss=0.2303, simple_loss=0.3069, pruned_loss=0.07683, over 8032.00 frames. ], tot_loss[loss=0.2342, simple_loss=0.3101, pruned_loss=0.0792, over 1622527.63 frames. ], batch size: 22, lr: 6.47e-03, grad_scale: 8.0 +2023-02-06 11:33:21,455 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90477.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:33:33,071 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=90492.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:33:48,752 INFO [train.py:901] (0/4) Epoch 12, batch 1600, loss[loss=0.2243, simple_loss=0.3143, pruned_loss=0.06713, over 8326.00 frames. ], tot_loss[loss=0.2325, simple_loss=0.3089, pruned_loss=0.07802, over 1618638.79 frames. ], batch size: 25, lr: 6.47e-03, grad_scale: 8.0 +2023-02-06 11:33:48,906 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 11:34:23,535 INFO [train.py:901] (0/4) Epoch 12, batch 1650, loss[loss=0.2427, simple_loss=0.3116, pruned_loss=0.08688, over 8134.00 frames. ], tot_loss[loss=0.2317, simple_loss=0.3082, pruned_loss=0.0776, over 1620098.72 frames. ], batch size: 22, lr: 6.46e-03, grad_scale: 8.0 +2023-02-06 11:34:43,449 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=90594.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:34:47,049 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90598.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:34:53,820 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90607.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:34:58,215 INFO [train.py:901] (0/4) Epoch 12, batch 1700, loss[loss=0.2766, simple_loss=0.3398, pruned_loss=0.1067, over 8040.00 frames. ], tot_loss[loss=0.231, simple_loss=0.3074, pruned_loss=0.07729, over 1616018.95 frames. ], batch size: 22, lr: 6.46e-03, grad_scale: 8.0 +2023-02-06 11:35:04,326 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90623.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:35:24,539 INFO [optim.py:369] (0/4) 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 11:35:32,824 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6583, 1.4955, 1.6215, 1.3948, 0.9172, 1.3925, 1.5365, 1.2877], + device='cuda:0'), covar=tensor([0.0512, 0.1214, 0.1647, 0.1335, 0.0548, 0.1472, 0.0700, 0.0655], + device='cuda:0'), in_proj_covar=tensor([0.0100, 0.0152, 0.0191, 0.0159, 0.0103, 0.0163, 0.0117, 0.0138], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 11:35:33,331 INFO [train.py:901] (0/4) Epoch 12, batch 1750, loss[loss=0.1947, simple_loss=0.2639, pruned_loss=0.0628, over 7410.00 frames. ], tot_loss[loss=0.2302, simple_loss=0.3069, pruned_loss=0.07672, over 1616560.72 frames. ], batch size: 17, lr: 6.46e-03, grad_scale: 8.0 +2023-02-06 11:35:40,293 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6939, 2.3723, 4.5745, 1.4171, 3.3946, 2.3603, 1.8960, 3.2307], + device='cuda:0'), covar=tensor([0.1718, 0.2329, 0.0627, 0.3882, 0.1313, 0.2738, 0.1789, 0.1960], + device='cuda:0'), in_proj_covar=tensor([0.0486, 0.0519, 0.0536, 0.0584, 0.0617, 0.0558, 0.0472, 0.0615], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 11:35:52,630 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0089, 1.2669, 1.2032, 0.6000, 1.2551, 1.0389, 0.0682, 1.1991], + device='cuda:0'), covar=tensor([0.0302, 0.0267, 0.0235, 0.0396, 0.0294, 0.0710, 0.0537, 0.0217], + device='cuda:0'), in_proj_covar=tensor([0.0391, 0.0332, 0.0277, 0.0384, 0.0315, 0.0471, 0.0356, 0.0356], + device='cuda:0'), out_proj_covar=tensor([1.1124e-04, 9.2123e-05, 7.7253e-05, 1.0763e-04, 8.8811e-05, 1.4259e-04, + 1.0116e-04, 1.0021e-04], device='cuda:0') +2023-02-06 11:35:55,167 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.6273, 1.8731, 4.6102, 1.8764, 2.6700, 5.2096, 5.1742, 4.5026], + device='cuda:0'), covar=tensor([0.1017, 0.1588, 0.0231, 0.2081, 0.0994, 0.0187, 0.0301, 0.0583], + device='cuda:0'), in_proj_covar=tensor([0.0265, 0.0296, 0.0261, 0.0292, 0.0277, 0.0236, 0.0348, 0.0292], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 11:36:04,029 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90709.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:36:07,408 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=90713.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:36:08,045 INFO [train.py:901] (0/4) Epoch 12, batch 1800, loss[loss=0.2408, simple_loss=0.3134, pruned_loss=0.08408, over 7525.00 frames. ], tot_loss[loss=0.2299, simple_loss=0.3068, pruned_loss=0.07647, over 1615778.69 frames. ], batch size: 18, lr: 6.46e-03, grad_scale: 8.0 +2023-02-06 11:36:28,905 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2926, 2.3789, 1.7177, 1.9968, 1.9530, 1.4322, 1.7302, 1.8535], + device='cuda:0'), covar=tensor([0.1373, 0.0385, 0.1109, 0.0631, 0.0643, 0.1366, 0.0968, 0.0881], + device='cuda:0'), in_proj_covar=tensor([0.0349, 0.0244, 0.0321, 0.0303, 0.0305, 0.0328, 0.0344, 0.0312], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 11:36:35,308 INFO [optim.py:369] (0/4) 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 11:36:43,319 INFO [train.py:901] (0/4) Epoch 12, batch 1850, loss[loss=0.2091, simple_loss=0.2894, pruned_loss=0.06437, over 7528.00 frames. ], tot_loss[loss=0.2304, simple_loss=0.3068, pruned_loss=0.077, over 1613045.23 frames. ], batch size: 18, lr: 6.46e-03, grad_scale: 8.0 +2023-02-06 11:37:17,706 INFO [train.py:901] (0/4) Epoch 12, batch 1900, loss[loss=0.2486, simple_loss=0.3348, pruned_loss=0.08117, over 8253.00 frames. ], tot_loss[loss=0.2304, simple_loss=0.3069, pruned_loss=0.07697, over 1611010.33 frames. ], batch size: 24, lr: 6.46e-03, grad_scale: 8.0 +2023-02-06 11:37:22,438 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=90821.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:37:27,319 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90828.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:37:38,160 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.94 vs. limit=5.0 +2023-02-06 11:37:44,436 INFO [optim.py:369] (0/4) 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 11:37:47,232 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-06 11:37:48,675 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=90858.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:37:50,836 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5779, 1.5450, 1.9888, 1.5348, 1.0260, 1.9861, 0.2327, 1.2949], + device='cuda:0'), covar=tensor([0.2519, 0.1776, 0.0583, 0.1669, 0.4296, 0.0577, 0.3449, 0.1930], + device='cuda:0'), in_proj_covar=tensor([0.0169, 0.0172, 0.0104, 0.0218, 0.0255, 0.0108, 0.0165, 0.0169], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 11:37:52,146 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90863.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:37:52,635 INFO [train.py:901] (0/4) Epoch 12, batch 1950, loss[loss=0.205, simple_loss=0.2692, pruned_loss=0.07043, over 7668.00 frames. ], tot_loss[loss=0.2314, simple_loss=0.3074, pruned_loss=0.07765, over 1609972.42 frames. ], batch size: 19, lr: 6.45e-03, grad_scale: 8.0 +2023-02-06 11:37:54,785 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90867.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:37:59,365 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-06 11:38:10,306 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90888.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:38:19,034 WARNING [train.py:1067] (0/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-06 11:38:27,979 INFO [train.py:901] (0/4) Epoch 12, batch 2000, loss[loss=0.2324, simple_loss=0.3182, pruned_loss=0.07332, over 8107.00 frames. ], tot_loss[loss=0.2313, simple_loss=0.3076, pruned_loss=0.07756, over 1611220.33 frames. ], batch size: 23, lr: 6.45e-03, grad_scale: 8.0 +2023-02-06 11:38:43,367 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90936.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:38:54,921 INFO [optim.py:369] (0/4) 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 11:39:02,893 INFO [train.py:901] (0/4) Epoch 12, batch 2050, loss[loss=0.1957, simple_loss=0.2618, pruned_loss=0.06482, over 7695.00 frames. ], tot_loss[loss=0.2319, simple_loss=0.308, pruned_loss=0.07788, over 1613798.78 frames. ], batch size: 18, lr: 6.45e-03, grad_scale: 8.0 +2023-02-06 11:39:03,755 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90965.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:39:09,912 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90973.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:39:21,818 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90990.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:39:38,672 INFO [train.py:901] (0/4) Epoch 12, batch 2100, loss[loss=0.2471, simple_loss=0.3214, pruned_loss=0.08637, over 8471.00 frames. ], tot_loss[loss=0.2328, simple_loss=0.3087, pruned_loss=0.07848, over 1612503.58 frames. ], batch size: 27, lr: 6.45e-03, grad_scale: 8.0 +2023-02-06 11:40:04,172 INFO [optim.py:369] (0/4) 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 11:40:12,107 INFO [train.py:901] (0/4) Epoch 12, batch 2150, loss[loss=0.2406, simple_loss=0.315, pruned_loss=0.08304, over 8548.00 frames. ], tot_loss[loss=0.2331, simple_loss=0.3086, pruned_loss=0.0788, over 1611283.68 frames. ], batch size: 39, lr: 6.45e-03, grad_scale: 8.0 +2023-02-06 11:40:26,820 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=91084.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:40:44,038 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91109.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:40:47,223 INFO [train.py:901] (0/4) Epoch 12, batch 2200, loss[loss=0.2067, simple_loss=0.2895, pruned_loss=0.06189, over 8122.00 frames. ], tot_loss[loss=0.2327, simple_loss=0.3079, pruned_loss=0.07878, over 1610642.85 frames. ], batch size: 22, lr: 6.44e-03, grad_scale: 8.0 +2023-02-06 11:41:13,716 INFO [optim.py:369] (0/4) 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 11:41:21,759 INFO [train.py:901] (0/4) Epoch 12, batch 2250, loss[loss=0.2588, simple_loss=0.3365, pruned_loss=0.09055, over 8327.00 frames. ], tot_loss[loss=0.2331, simple_loss=0.3079, pruned_loss=0.0791, over 1603866.01 frames. ], batch size: 26, lr: 6.44e-03, grad_scale: 8.0 +2023-02-06 11:41:41,120 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=91192.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:41:54,520 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=91211.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:41:56,410 INFO [train.py:901] (0/4) Epoch 12, batch 2300, loss[loss=0.2309, simple_loss=0.2933, pruned_loss=0.0842, over 7801.00 frames. ], tot_loss[loss=0.2337, simple_loss=0.3087, pruned_loss=0.07934, over 1606119.60 frames. ], batch size: 19, lr: 6.44e-03, grad_scale: 8.0 +2023-02-06 11:41:58,501 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91217.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:42:07,351 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=91229.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:42:23,426 INFO [optim.py:369] (0/4) 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 11:42:25,000 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91254.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:42:31,680 INFO [train.py:901] (0/4) Epoch 12, batch 2350, loss[loss=0.2682, simple_loss=0.3237, pruned_loss=0.1063, over 7932.00 frames. ], tot_loss[loss=0.2329, simple_loss=0.3087, pruned_loss=0.07852, over 1614400.78 frames. ], batch size: 20, lr: 6.44e-03, grad_scale: 8.0 +2023-02-06 11:42:57,733 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=91303.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:43:05,701 INFO [train.py:901] (0/4) Epoch 12, batch 2400, loss[loss=0.219, simple_loss=0.2974, pruned_loss=0.07032, over 8341.00 frames. ], tot_loss[loss=0.2318, simple_loss=0.3077, pruned_loss=0.07798, over 1615081.69 frames. ], batch size: 26, lr: 6.44e-03, grad_scale: 8.0 +2023-02-06 11:43:14,306 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91326.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:43:32,241 INFO [optim.py:369] (0/4) 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 11:43:35,834 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.2118, 1.1589, 3.3539, 1.0329, 2.9513, 2.8045, 3.0810, 2.9459], + device='cuda:0'), covar=tensor([0.0688, 0.3751, 0.0707, 0.3337, 0.1252, 0.0979, 0.0645, 0.0831], + device='cuda:0'), in_proj_covar=tensor([0.0498, 0.0580, 0.0588, 0.0536, 0.0609, 0.0524, 0.0518, 0.0583], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 11:43:41,049 INFO [train.py:901] (0/4) Epoch 12, batch 2450, loss[loss=0.2841, simple_loss=0.3465, pruned_loss=0.1108, over 8502.00 frames. ], tot_loss[loss=0.2325, simple_loss=0.3082, pruned_loss=0.07836, over 1614623.38 frames. ], batch size: 26, lr: 6.44e-03, grad_scale: 8.0 +2023-02-06 11:44:06,928 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7437, 1.9931, 2.1687, 1.3749, 2.3245, 1.3791, 0.7723, 1.9012], + device='cuda:0'), covar=tensor([0.0471, 0.0266, 0.0194, 0.0418, 0.0247, 0.0704, 0.0602, 0.0230], + device='cuda:0'), in_proj_covar=tensor([0.0397, 0.0332, 0.0280, 0.0384, 0.0317, 0.0473, 0.0362, 0.0359], + device='cuda:0'), out_proj_covar=tensor([1.1284e-04, 9.1987e-05, 7.7832e-05, 1.0725e-04, 8.9322e-05, 1.4282e-04, + 1.0273e-04, 1.0109e-04], device='cuda:0') +2023-02-06 11:44:11,811 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3595, 2.0049, 2.8729, 2.2937, 2.6425, 2.1947, 1.7818, 1.3710], + device='cuda:0'), covar=tensor([0.3719, 0.3774, 0.1163, 0.2543, 0.1679, 0.2146, 0.1725, 0.4005], + device='cuda:0'), in_proj_covar=tensor([0.0883, 0.0871, 0.0740, 0.0850, 0.0937, 0.0804, 0.0704, 0.0765], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 11:44:15,054 INFO [train.py:901] (0/4) Epoch 12, batch 2500, loss[loss=0.1911, simple_loss=0.262, pruned_loss=0.06007, over 7800.00 frames. ], tot_loss[loss=0.2319, simple_loss=0.308, pruned_loss=0.0779, over 1615724.75 frames. ], batch size: 19, lr: 6.43e-03, grad_scale: 8.0 +2023-02-06 11:44:41,745 INFO [optim.py:369] (0/4) 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 11:44:50,272 INFO [train.py:901] (0/4) Epoch 12, batch 2550, loss[loss=0.2593, simple_loss=0.3304, pruned_loss=0.09411, over 8488.00 frames. ], tot_loss[loss=0.2328, simple_loss=0.3088, pruned_loss=0.07836, over 1622332.73 frames. ], batch size: 29, lr: 6.43e-03, grad_scale: 8.0 +2023-02-06 11:45:24,410 INFO [train.py:901] (0/4) Epoch 12, batch 2600, loss[loss=0.2348, simple_loss=0.3076, pruned_loss=0.08104, over 8370.00 frames. ], tot_loss[loss=0.2327, simple_loss=0.3091, pruned_loss=0.07819, over 1621174.42 frames. ], batch size: 24, lr: 6.43e-03, grad_scale: 8.0 +2023-02-06 11:45:50,006 INFO [optim.py:369] (0/4) 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 11:45:56,363 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=91560.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 11:45:58,890 INFO [train.py:901] (0/4) Epoch 12, batch 2650, loss[loss=0.2152, simple_loss=0.2912, pruned_loss=0.06963, over 7930.00 frames. ], tot_loss[loss=0.2332, simple_loss=0.3088, pruned_loss=0.07875, over 1616344.70 frames. ], batch size: 20, lr: 6.43e-03, grad_scale: 8.0 +2023-02-06 11:46:11,887 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=91582.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:46:29,366 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91607.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:46:33,843 INFO [train.py:901] (0/4) Epoch 12, batch 2700, loss[loss=0.2246, simple_loss=0.3037, pruned_loss=0.0728, over 8484.00 frames. ], tot_loss[loss=0.232, simple_loss=0.308, pruned_loss=0.07799, over 1616490.11 frames. ], batch size: 26, lr: 6.43e-03, grad_scale: 8.0 +2023-02-06 11:46:35,603 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 +2023-02-06 11:46:55,919 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=91647.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:46:59,281 INFO [optim.py:369] (0/4) 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 11:47:08,024 INFO [train.py:901] (0/4) Epoch 12, batch 2750, loss[loss=0.2255, simple_loss=0.3117, pruned_loss=0.06961, over 8443.00 frames. ], tot_loss[loss=0.231, simple_loss=0.307, pruned_loss=0.07748, over 1614240.70 frames. ], batch size: 24, lr: 6.43e-03, grad_scale: 8.0 +2023-02-06 11:47:43,511 INFO [train.py:901] (0/4) Epoch 12, batch 2800, loss[loss=0.2218, simple_loss=0.2878, pruned_loss=0.07793, over 7649.00 frames. ], tot_loss[loss=0.2322, simple_loss=0.3076, pruned_loss=0.07841, over 1614779.33 frames. ], batch size: 19, lr: 6.42e-03, grad_scale: 8.0 +2023-02-06 11:48:08,849 INFO [optim.py:369] (0/4) 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 11:48:13,228 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 +2023-02-06 11:48:15,796 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 12, batch 2850, loss[loss=0.2538, simple_loss=0.3298, pruned_loss=0.08889, over 8456.00 frames. ], tot_loss[loss=0.2321, simple_loss=0.3076, pruned_loss=0.07831, over 1612668.95 frames. ], batch size: 27, lr: 6.42e-03, grad_scale: 8.0 +2023-02-06 11:48:36,284 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 +2023-02-06 11:48:52,949 INFO [train.py:901] (0/4) Epoch 12, batch 2900, loss[loss=0.2531, simple_loss=0.332, pruned_loss=0.08712, over 8519.00 frames. ], tot_loss[loss=0.2316, simple_loss=0.3072, pruned_loss=0.07804, over 1612778.92 frames. ], batch size: 28, lr: 6.42e-03, grad_scale: 8.0 +2023-02-06 11:49:18,828 INFO [optim.py:369] (0/4) 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] (0/4) 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] (0/4) Epoch 12, batch 2950, loss[loss=0.2358, simple_loss=0.313, pruned_loss=0.07934, over 8097.00 frames. ], tot_loss[loss=0.2323, simple_loss=0.308, pruned_loss=0.07833, over 1614757.16 frames. ], batch size: 23, lr: 6.42e-03, grad_scale: 8.0 +2023-02-06 11:49:49,030 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.58 vs. limit=2.0 +2023-02-06 11:49:54,040 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=91904.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 11:49:57,389 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=91909.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:50:00,612 INFO [train.py:901] (0/4) Epoch 12, batch 3000, loss[loss=0.2315, simple_loss=0.3089, pruned_loss=0.077, over 8603.00 frames. ], tot_loss[loss=0.2319, simple_loss=0.3073, pruned_loss=0.07823, over 1613887.24 frames. ], batch size: 39, lr: 6.42e-03, grad_scale: 8.0 +2023-02-06 11:50:00,613 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 11:50:12,223 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.4731, 3.3838, 3.1917, 2.0220, 3.1259, 3.0664, 3.2922, 2.7886], + device='cuda:0'), covar=tensor([0.0918, 0.0710, 0.0819, 0.4092, 0.0798, 0.1154, 0.1050, 0.1121], + device='cuda:0'), in_proj_covar=tensor([0.0466, 0.0379, 0.0388, 0.0489, 0.0379, 0.0383, 0.0379, 0.0330], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 11:50:13,629 INFO [train.py:935] (0/4) Epoch 12, validation: loss=0.1868, simple_loss=0.2871, pruned_loss=0.04323, over 944034.00 frames. +2023-02-06 11:50:13,630 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-06 11:50:40,670 INFO [optim.py:369] (0/4) 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 11:50:49,095 INFO [train.py:901] (0/4) Epoch 12, batch 3050, loss[loss=0.2566, simple_loss=0.3255, pruned_loss=0.0939, over 8454.00 frames. ], tot_loss[loss=0.2331, simple_loss=0.3083, pruned_loss=0.07891, over 1617585.74 frames. ], batch size: 29, lr: 6.41e-03, grad_scale: 8.0 +2023-02-06 11:50:56,129 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=91973.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:51:10,055 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3344, 2.6368, 2.9362, 1.4621, 3.1723, 1.7303, 1.5247, 2.1174], + device='cuda:0'), covar=tensor([0.0558, 0.0269, 0.0217, 0.0553, 0.0330, 0.0663, 0.0649, 0.0322], + device='cuda:0'), in_proj_covar=tensor([0.0395, 0.0331, 0.0280, 0.0387, 0.0320, 0.0475, 0.0358, 0.0357], + device='cuda:0'), out_proj_covar=tensor([1.1218e-04, 9.1945e-05, 7.7681e-05, 1.0791e-04, 8.9865e-05, 1.4358e-04, + 1.0141e-04, 1.0051e-04], device='cuda:0') +2023-02-06 11:51:14,055 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=91999.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 11:51:14,720 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-92000.pt +2023-02-06 11:51:25,121 INFO [train.py:901] (0/4) Epoch 12, batch 3100, loss[loss=0.2058, simple_loss=0.296, pruned_loss=0.05785, over 8107.00 frames. ], tot_loss[loss=0.2319, simple_loss=0.3077, pruned_loss=0.07808, over 1621148.75 frames. ], batch size: 23, lr: 6.41e-03, grad_scale: 8.0 +2023-02-06 11:51:28,135 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92018.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:51:28,802 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92019.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 11:51:45,803 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92043.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:51:51,735 INFO [optim.py:369] (0/4) 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 11:52:01,129 INFO [train.py:901] (0/4) Epoch 12, batch 3150, loss[loss=0.3255, simple_loss=0.3632, pruned_loss=0.1438, over 6917.00 frames. ], tot_loss[loss=0.2319, simple_loss=0.3076, pruned_loss=0.07813, over 1620680.47 frames. ], batch size: 71, lr: 6.41e-03, grad_scale: 16.0 +2023-02-06 11:52:18,313 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.7003, 1.5176, 4.8422, 1.7280, 4.3213, 3.9543, 4.4011, 4.2127], + device='cuda:0'), covar=tensor([0.0473, 0.4188, 0.0386, 0.3503, 0.0998, 0.0804, 0.0508, 0.0581], + device='cuda:0'), in_proj_covar=tensor([0.0500, 0.0577, 0.0591, 0.0539, 0.0619, 0.0528, 0.0521, 0.0586], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 11:52:35,767 INFO [train.py:901] (0/4) Epoch 12, batch 3200, loss[loss=0.2071, simple_loss=0.2941, pruned_loss=0.06011, over 8356.00 frames. ], tot_loss[loss=0.2313, simple_loss=0.3071, pruned_loss=0.07777, over 1614115.28 frames. ], batch size: 24, lr: 6.41e-03, grad_scale: 16.0 +2023-02-06 11:52:44,075 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.23 vs. limit=5.0 +2023-02-06 11:53:02,007 INFO [optim.py:369] (0/4) 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 11:53:10,360 INFO [train.py:901] (0/4) Epoch 12, batch 3250, loss[loss=0.1909, simple_loss=0.2657, pruned_loss=0.05804, over 7546.00 frames. ], tot_loss[loss=0.2318, simple_loss=0.3074, pruned_loss=0.07815, over 1612086.93 frames. ], batch size: 18, lr: 6.41e-03, grad_scale: 16.0 +2023-02-06 11:53:46,146 INFO [train.py:901] (0/4) Epoch 12, batch 3300, loss[loss=0.2064, simple_loss=0.2934, pruned_loss=0.05973, over 8359.00 frames. ], tot_loss[loss=0.2308, simple_loss=0.307, pruned_loss=0.07731, over 1615094.69 frames. ], batch size: 24, lr: 6.41e-03, grad_scale: 16.0 +2023-02-06 11:54:11,040 INFO [optim.py:369] (0/4) 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 11:54:11,765 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=92253.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:54:19,116 INFO [train.py:901] (0/4) Epoch 12, batch 3350, loss[loss=0.2323, simple_loss=0.3142, pruned_loss=0.07517, over 7816.00 frames. ], tot_loss[loss=0.2325, simple_loss=0.3091, pruned_loss=0.07794, over 1620060.59 frames. ], batch size: 20, lr: 6.40e-03, grad_scale: 16.0 +2023-02-06 11:54:27,369 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92275.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 11:54:45,189 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92300.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 11:54:55,343 INFO [train.py:901] (0/4) Epoch 12, batch 3400, loss[loss=0.1989, simple_loss=0.2894, pruned_loss=0.05417, over 8324.00 frames. ], tot_loss[loss=0.2318, simple_loss=0.3084, pruned_loss=0.07754, over 1616294.59 frames. ], batch size: 25, lr: 6.40e-03, grad_scale: 16.0 +2023-02-06 11:54:57,470 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=92317.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:55:15,992 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=92343.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 11:55:21,867 INFO [optim.py:369] (0/4) 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 11:55:26,204 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.9165, 2.6896, 3.6363, 1.9654, 1.7817, 3.5269, 0.7636, 2.0441], + device='cuda:0'), covar=tensor([0.2566, 0.1539, 0.0298, 0.3065, 0.4447, 0.0393, 0.4056, 0.2457], + device='cuda:0'), in_proj_covar=tensor([0.0172, 0.0175, 0.0104, 0.0221, 0.0259, 0.0109, 0.0167, 0.0169], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 11:55:26,341 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.47 vs. limit=5.0 +2023-02-06 11:55:29,930 INFO [train.py:901] (0/4) Epoch 12, batch 3450, loss[loss=0.25, simple_loss=0.3268, pruned_loss=0.08665, over 8240.00 frames. ], tot_loss[loss=0.2319, simple_loss=0.3087, pruned_loss=0.07757, over 1620254.31 frames. ], batch size: 24, lr: 6.40e-03, grad_scale: 16.0 +2023-02-06 11:55:32,853 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92368.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:55:34,186 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2102, 1.8802, 2.6169, 2.2648, 2.5461, 2.1290, 1.7228, 1.1745], + device='cuda:0'), covar=tensor([0.3797, 0.3675, 0.1343, 0.2473, 0.1771, 0.2289, 0.1652, 0.4021], + device='cuda:0'), in_proj_covar=tensor([0.0890, 0.0870, 0.0732, 0.0845, 0.0934, 0.0804, 0.0705, 0.0766], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 11:55:52,772 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 +2023-02-06 11:56:04,892 INFO [train.py:901] (0/4) Epoch 12, batch 3500, loss[loss=0.2637, simple_loss=0.329, pruned_loss=0.09927, over 7979.00 frames. ], tot_loss[loss=0.2314, simple_loss=0.3081, pruned_loss=0.07737, over 1615880.29 frames. ], batch size: 21, lr: 6.40e-03, grad_scale: 8.0 +2023-02-06 11:56:18,391 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92432.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:56:29,126 WARNING [train.py:1067] (0/4) 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] (0/4) 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 11:56:36,511 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92458.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 11:56:40,246 INFO [train.py:901] (0/4) Epoch 12, batch 3550, loss[loss=0.2371, simple_loss=0.3097, pruned_loss=0.08219, over 7817.00 frames. ], tot_loss[loss=0.2321, simple_loss=0.3088, pruned_loss=0.07772, over 1617618.50 frames. ], batch size: 20, lr: 6.40e-03, grad_scale: 8.0 +2023-02-06 11:57:11,660 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 +2023-02-06 11:57:14,698 INFO [train.py:901] (0/4) Epoch 12, batch 3600, loss[loss=0.1834, simple_loss=0.2688, pruned_loss=0.04899, over 8191.00 frames. ], tot_loss[loss=0.232, simple_loss=0.3089, pruned_loss=0.07753, over 1615842.61 frames. ], batch size: 23, lr: 6.40e-03, grad_scale: 8.0 +2023-02-06 11:57:29,923 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.74 vs. limit=5.0 +2023-02-06 11:57:38,956 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.08 vs. limit=5.0 +2023-02-06 11:57:42,450 INFO [optim.py:369] (0/4) 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 11:57:50,891 INFO [train.py:901] (0/4) Epoch 12, batch 3650, loss[loss=0.2198, simple_loss=0.2862, pruned_loss=0.0767, over 7926.00 frames. ], tot_loss[loss=0.2312, simple_loss=0.3073, pruned_loss=0.07752, over 1610797.73 frames. ], batch size: 20, lr: 6.39e-03, grad_scale: 8.0 +2023-02-06 11:58:10,859 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.6648, 4.6991, 4.2023, 1.9275, 4.1947, 4.1932, 4.2899, 3.9743], + device='cuda:0'), covar=tensor([0.0668, 0.0463, 0.0887, 0.4793, 0.0710, 0.0911, 0.1161, 0.0829], + device='cuda:0'), in_proj_covar=tensor([0.0468, 0.0377, 0.0388, 0.0488, 0.0380, 0.0383, 0.0379, 0.0332], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 11:58:23,913 INFO [train.py:901] (0/4) Epoch 12, batch 3700, loss[loss=0.2766, simple_loss=0.3331, pruned_loss=0.11, over 7928.00 frames. ], tot_loss[loss=0.2331, simple_loss=0.3092, pruned_loss=0.07847, over 1615997.65 frames. ], batch size: 20, lr: 6.39e-03, grad_scale: 8.0 +2023-02-06 11:58:28,530 WARNING [train.py:1067] (0/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-06 11:58:30,697 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92624.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:58:44,254 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92643.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:58:48,471 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92649.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:58:50,889 INFO [optim.py:369] (0/4) 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 11:58:54,981 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.5852, 1.6264, 4.3231, 1.8238, 2.4406, 4.9399, 4.9791, 4.3211], + device='cuda:0'), covar=tensor([0.1021, 0.1734, 0.0272, 0.2246, 0.1159, 0.0208, 0.0313, 0.0582], + device='cuda:0'), in_proj_covar=tensor([0.0270, 0.0302, 0.0264, 0.0298, 0.0280, 0.0240, 0.0354, 0.0295], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 11:58:57,877 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3604, 1.9236, 2.9064, 2.2573, 2.6943, 2.2139, 1.8086, 1.4340], + device='cuda:0'), covar=tensor([0.4204, 0.4418, 0.1249, 0.2897, 0.2039, 0.2344, 0.1847, 0.4294], + device='cuda:0'), in_proj_covar=tensor([0.0896, 0.0874, 0.0731, 0.0854, 0.0939, 0.0809, 0.0706, 0.0770], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 11:58:59,706 INFO [train.py:901] (0/4) Epoch 12, batch 3750, loss[loss=0.1965, simple_loss=0.2686, pruned_loss=0.0622, over 7525.00 frames. ], tot_loss[loss=0.2312, simple_loss=0.3074, pruned_loss=0.07751, over 1612350.31 frames. ], batch size: 18, lr: 6.39e-03, grad_scale: 8.0 +2023-02-06 11:59:01,856 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.3704, 1.6275, 4.5252, 1.6316, 4.0526, 3.7498, 4.0907, 3.9773], + device='cuda:0'), covar=tensor([0.0500, 0.3974, 0.0521, 0.3525, 0.1115, 0.0909, 0.0517, 0.0594], + device='cuda:0'), in_proj_covar=tensor([0.0503, 0.0579, 0.0600, 0.0540, 0.0620, 0.0533, 0.0524, 0.0585], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 11:59:17,063 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92688.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:59:28,223 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6903, 2.0064, 2.2373, 1.1849, 2.3289, 1.4123, 0.7050, 1.7075], + device='cuda:0'), covar=tensor([0.0451, 0.0224, 0.0162, 0.0443, 0.0245, 0.0706, 0.0594, 0.0263], + device='cuda:0'), in_proj_covar=tensor([0.0408, 0.0337, 0.0286, 0.0400, 0.0328, 0.0489, 0.0365, 0.0367], + device='cuda:0'), 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:0') +2023-02-06 11:59:34,439 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 12, batch 3800, loss[loss=0.2354, simple_loss=0.3209, pruned_loss=0.07497, over 8597.00 frames. ], tot_loss[loss=0.2299, simple_loss=0.306, pruned_loss=0.0769, over 1613108.52 frames. ], batch size: 31, lr: 6.39e-03, grad_scale: 8.0 +2023-02-06 11:59:35,188 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92714.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 11:59:52,016 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92738.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 11:59:52,764 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92739.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 12:00:02,136 INFO [optim.py:369] (0/4) 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 12:00:09,493 INFO [train.py:901] (0/4) Epoch 12, batch 3850, loss[loss=0.2023, simple_loss=0.284, pruned_loss=0.06026, over 8584.00 frames. ], tot_loss[loss=0.2303, simple_loss=0.3065, pruned_loss=0.07706, over 1612521.94 frames. ], batch size: 31, lr: 6.39e-03, grad_scale: 8.0 +2023-02-06 12:00:33,580 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-06 12:00:45,119 INFO [train.py:901] (0/4) Epoch 12, batch 3900, loss[loss=0.1977, simple_loss=0.2804, pruned_loss=0.05745, over 7816.00 frames. ], tot_loss[loss=0.2298, simple_loss=0.3059, pruned_loss=0.0769, over 1610584.35 frames. ], batch size: 20, lr: 6.39e-03, grad_scale: 8.0 +2023-02-06 12:01:08,848 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 12:01:17,562 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0687, 2.2371, 1.7953, 2.9001, 1.4964, 1.5247, 2.0388, 2.3947], + device='cuda:0'), covar=tensor([0.0736, 0.0928, 0.0997, 0.0419, 0.1141, 0.1507, 0.1047, 0.0696], + device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0212, 0.0254, 0.0215, 0.0216, 0.0251, 0.0256, 0.0216], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 12:01:19,448 INFO [train.py:901] (0/4) Epoch 12, batch 3950, loss[loss=0.1863, simple_loss=0.2637, pruned_loss=0.05446, over 7791.00 frames. ], tot_loss[loss=0.2299, simple_loss=0.3058, pruned_loss=0.07699, over 1607390.17 frames. ], batch size: 19, lr: 6.38e-03, grad_scale: 8.0 +2023-02-06 12:01:53,518 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 +2023-02-06 12:01:54,553 INFO [train.py:901] (0/4) Epoch 12, batch 4000, loss[loss=0.2408, simple_loss=0.3244, pruned_loss=0.07863, over 8534.00 frames. ], tot_loss[loss=0.23, simple_loss=0.3062, pruned_loss=0.07688, over 1613355.69 frames. ], batch size: 49, lr: 6.38e-03, grad_scale: 8.0 +2023-02-06 12:01:56,834 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92917.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:02:00,211 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.8475, 3.8105, 3.4860, 1.6710, 3.4198, 3.3622, 3.5003, 3.2162], + device='cuda:0'), covar=tensor([0.0863, 0.0643, 0.0999, 0.4671, 0.0816, 0.1112, 0.1255, 0.0869], + device='cuda:0'), in_proj_covar=tensor([0.0463, 0.0376, 0.0386, 0.0483, 0.0376, 0.0380, 0.0374, 0.0332], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 12:02:05,009 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7231, 2.0231, 3.2634, 1.4657, 2.4657, 1.9910, 1.8006, 2.2985], + device='cuda:0'), covar=tensor([0.1584, 0.2202, 0.0697, 0.3683, 0.1436, 0.2775, 0.1759, 0.2194], + device='cuda:0'), in_proj_covar=tensor([0.0489, 0.0527, 0.0538, 0.0581, 0.0622, 0.0560, 0.0476, 0.0614], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 12:02:08,139 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0733, 1.6871, 3.4421, 1.4696, 2.4026, 3.8414, 3.8557, 3.2444], + device='cuda:0'), covar=tensor([0.1095, 0.1515, 0.0350, 0.2158, 0.1032, 0.0238, 0.0436, 0.0636], + device='cuda:0'), in_proj_covar=tensor([0.0270, 0.0301, 0.0265, 0.0297, 0.0278, 0.0241, 0.0354, 0.0298], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 12:02:18,336 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92949.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:02:20,901 INFO [optim.py:369] (0/4) 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 12:02:28,380 INFO [train.py:901] (0/4) Epoch 12, batch 4050, loss[loss=0.2243, simple_loss=0.304, pruned_loss=0.07231, over 8458.00 frames. ], tot_loss[loss=0.2315, simple_loss=0.3077, pruned_loss=0.07765, over 1616211.13 frames. ], batch size: 27, lr: 6.38e-03, grad_scale: 8.0 +2023-02-06 12:02:44,175 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=92987.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:02:48,258 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92993.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:02:57,115 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1859, 2.1691, 2.4804, 1.9157, 1.8163, 2.4727, 1.1540, 2.0084], + device='cuda:0'), covar=tensor([0.2735, 0.1413, 0.0460, 0.1843, 0.2620, 0.0540, 0.2842, 0.1655], + device='cuda:0'), in_proj_covar=tensor([0.0170, 0.0174, 0.0103, 0.0216, 0.0253, 0.0107, 0.0162, 0.0168], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 12:03:03,714 INFO [train.py:901] (0/4) Epoch 12, batch 4100, loss[loss=0.2374, simple_loss=0.3102, pruned_loss=0.08232, over 8142.00 frames. ], tot_loss[loss=0.2326, simple_loss=0.3085, pruned_loss=0.07836, over 1614351.40 frames. ], batch size: 22, lr: 6.38e-03, grad_scale: 8.0 +2023-02-06 12:03:04,548 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0141, 1.4935, 1.5883, 1.2104, 0.9556, 1.3914, 1.8104, 1.8895], + device='cuda:0'), covar=tensor([0.0485, 0.1326, 0.1762, 0.1481, 0.0617, 0.1534, 0.0673, 0.0521], + device='cuda:0'), in_proj_covar=tensor([0.0100, 0.0153, 0.0194, 0.0160, 0.0104, 0.0164, 0.0117, 0.0139], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0008, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 12:03:13,897 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93028.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:03:21,022 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 +2023-02-06 12:03:30,615 INFO [optim.py:369] (0/4) 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 12:03:31,415 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 12, batch 4150, loss[loss=0.1978, simple_loss=0.2775, pruned_loss=0.05905, over 7810.00 frames. ], tot_loss[loss=0.2313, simple_loss=0.3076, pruned_loss=0.07746, over 1616789.49 frames. ], batch size: 20, lr: 6.38e-03, grad_scale: 8.0 +2023-02-06 12:03:50,877 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=93082.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:04:04,580 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93102.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:04:05,524 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 +2023-02-06 12:04:12,248 INFO [train.py:901] (0/4) Epoch 12, batch 4200, loss[loss=0.2144, simple_loss=0.2909, pruned_loss=0.06897, over 7532.00 frames. ], tot_loss[loss=0.2306, simple_loss=0.3068, pruned_loss=0.07719, over 1613237.01 frames. ], batch size: 18, lr: 6.38e-03, grad_scale: 8.0 +2023-02-06 12:04:13,210 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.6821, 2.1774, 3.5582, 2.5175, 3.0640, 2.3520, 2.0125, 1.6666], + device='cuda:0'), covar=tensor([0.4031, 0.4715, 0.1271, 0.3313, 0.2153, 0.2454, 0.1737, 0.4664], + device='cuda:0'), in_proj_covar=tensor([0.0885, 0.0864, 0.0720, 0.0845, 0.0927, 0.0796, 0.0699, 0.0758], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 12:04:24,925 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-06 12:04:26,408 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93133.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:04:29,266 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8817, 1.9063, 2.3323, 1.9461, 1.3897, 2.3789, 0.3764, 1.4391], + device='cuda:0'), covar=tensor([0.2648, 0.1860, 0.0490, 0.1553, 0.4002, 0.0493, 0.3469, 0.2261], + device='cuda:0'), in_proj_covar=tensor([0.0169, 0.0172, 0.0101, 0.0214, 0.0252, 0.0107, 0.0161, 0.0167], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 12:04:40,123 INFO [optim.py:369] (0/4) 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 12:04:47,442 INFO [train.py:901] (0/4) Epoch 12, batch 4250, loss[loss=0.2447, simple_loss=0.3221, pruned_loss=0.08364, over 8248.00 frames. ], tot_loss[loss=0.2308, simple_loss=0.307, pruned_loss=0.07727, over 1613399.47 frames. ], batch size: 24, lr: 6.37e-03, grad_scale: 8.0 +2023-02-06 12:04:48,809 WARNING [train.py:1067] (0/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-06 12:04:53,107 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.81 vs. limit=2.0 +2023-02-06 12:05:06,762 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=93193.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:05:09,364 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93197.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:05:21,310 INFO [train.py:901] (0/4) Epoch 12, batch 4300, loss[loss=0.2347, simple_loss=0.315, pruned_loss=0.07721, over 8354.00 frames. ], tot_loss[loss=0.2309, simple_loss=0.3071, pruned_loss=0.07733, over 1616374.95 frames. ], batch size: 24, lr: 6.37e-03, grad_scale: 8.0 +2023-02-06 12:05:22,134 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.7083, 1.3910, 4.8674, 1.8032, 4.3047, 4.0463, 4.3802, 4.2195], + device='cuda:0'), covar=tensor([0.0454, 0.4166, 0.0377, 0.3226, 0.0946, 0.0762, 0.0480, 0.0528], + device='cuda:0'), in_proj_covar=tensor([0.0501, 0.0580, 0.0594, 0.0537, 0.0615, 0.0527, 0.0519, 0.0581], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 12:05:48,576 INFO [optim.py:369] (0/4) 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 12:05:54,511 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 12, batch 4350, loss[loss=0.1719, simple_loss=0.2508, pruned_loss=0.0465, over 7556.00 frames. ], tot_loss[loss=0.2308, simple_loss=0.307, pruned_loss=0.07732, over 1615280.72 frames. ], batch size: 18, lr: 6.37e-03, grad_scale: 8.0 +2023-02-06 12:06:15,789 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=93293.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:06:16,409 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-06 12:06:25,902 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93308.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:06:29,907 INFO [train.py:901] (0/4) Epoch 12, batch 4400, loss[loss=0.271, simple_loss=0.3431, pruned_loss=0.09943, over 8634.00 frames. ], tot_loss[loss=0.2316, simple_loss=0.3076, pruned_loss=0.07775, over 1617587.66 frames. ], batch size: 34, lr: 6.37e-03, grad_scale: 8.0 +2023-02-06 12:06:46,133 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=93337.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:06:51,665 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1819, 3.0360, 2.3539, 2.5631, 2.5119, 2.1201, 2.4064, 2.7794], + device='cuda:0'), covar=tensor([0.1023, 0.0306, 0.0731, 0.0505, 0.0522, 0.0935, 0.0765, 0.0674], + device='cuda:0'), in_proj_covar=tensor([0.0341, 0.0234, 0.0313, 0.0296, 0.0297, 0.0323, 0.0341, 0.0301], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 12:06:58,349 INFO [optim.py:369] (0/4) 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 12:06:58,369 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-06 12:07:01,786 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 12, batch 4450, loss[loss=0.1936, simple_loss=0.2761, pruned_loss=0.05554, over 8250.00 frames. ], tot_loss[loss=0.2309, simple_loss=0.3069, pruned_loss=0.07744, over 1611604.42 frames. ], batch size: 24, lr: 6.37e-03, grad_scale: 8.0 +2023-02-06 12:07:11,753 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=93372.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:07:14,525 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93376.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:07:19,341 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93383.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:07:29,365 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=93398.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:07:36,045 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93408.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:07:39,924 INFO [train.py:901] (0/4) Epoch 12, batch 4500, loss[loss=0.2254, simple_loss=0.308, pruned_loss=0.07138, over 8362.00 frames. ], tot_loss[loss=0.2313, simple_loss=0.3071, pruned_loss=0.07781, over 1609764.44 frames. ], batch size: 26, lr: 6.37e-03, grad_scale: 8.0 +2023-02-06 12:07:46,172 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1343, 3.9613, 2.5466, 2.9374, 2.8149, 1.9526, 2.6589, 3.0825], + device='cuda:0'), covar=tensor([0.1333, 0.0274, 0.0823, 0.0596, 0.0646, 0.1153, 0.0924, 0.0923], + device='cuda:0'), in_proj_covar=tensor([0.0339, 0.0231, 0.0310, 0.0293, 0.0295, 0.0320, 0.0337, 0.0298], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 12:07:50,706 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-06 12:08:01,527 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3985, 2.3836, 1.7748, 2.0541, 1.8913, 1.4026, 1.7485, 1.8343], + device='cuda:0'), covar=tensor([0.1150, 0.0317, 0.0863, 0.0465, 0.0630, 0.1177, 0.0923, 0.0768], + device='cuda:0'), in_proj_covar=tensor([0.0340, 0.0233, 0.0312, 0.0294, 0.0297, 0.0321, 0.0338, 0.0299], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 12:08:06,004 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93452.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:08:06,469 INFO [optim.py:369] (0/4) 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 12:08:06,701 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93453.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:08:13,859 INFO [train.py:901] (0/4) Epoch 12, batch 4550, loss[loss=0.2088, simple_loss=0.2974, pruned_loss=0.06007, over 8476.00 frames. ], tot_loss[loss=0.2308, simple_loss=0.3066, pruned_loss=0.07744, over 1607504.76 frames. ], batch size: 29, lr: 6.36e-03, grad_scale: 8.0 +2023-02-06 12:08:24,170 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=93477.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:08:24,981 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93478.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:08:30,575 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9284, 2.3667, 4.5017, 1.5595, 3.1011, 2.5059, 2.0343, 2.9642], + device='cuda:0'), covar=tensor([0.1533, 0.2261, 0.0701, 0.3710, 0.1524, 0.2373, 0.1679, 0.2193], + device='cuda:0'), in_proj_covar=tensor([0.0485, 0.0523, 0.0530, 0.0577, 0.0613, 0.0549, 0.0472, 0.0611], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 12:08:31,876 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93487.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:08:49,629 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93513.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:08:50,182 INFO [train.py:901] (0/4) Epoch 12, batch 4600, loss[loss=0.2329, simple_loss=0.3034, pruned_loss=0.08117, over 8702.00 frames. ], tot_loss[loss=0.23, simple_loss=0.306, pruned_loss=0.07695, over 1607968.85 frames. ], batch size: 34, lr: 6.36e-03, grad_scale: 8.0 +2023-02-06 12:09:16,605 INFO [optim.py:369] (0/4) 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 12:09:21,186 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 +2023-02-06 12:09:24,867 INFO [train.py:901] (0/4) Epoch 12, batch 4650, loss[loss=0.2611, simple_loss=0.3402, pruned_loss=0.09099, over 8188.00 frames. ], tot_loss[loss=0.2302, simple_loss=0.3061, pruned_loss=0.07716, over 1609704.26 frames. ], batch size: 23, lr: 6.36e-03, grad_scale: 8.0 +2023-02-06 12:09:25,070 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93564.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:09:41,891 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93589.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:09:45,161 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93592.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:09:59,875 INFO [train.py:901] (0/4) Epoch 12, batch 4700, loss[loss=0.2283, simple_loss=0.3031, pruned_loss=0.07676, over 7658.00 frames. ], tot_loss[loss=0.229, simple_loss=0.3049, pruned_loss=0.07656, over 1605649.06 frames. ], batch size: 19, lr: 6.36e-03, grad_scale: 8.0 +2023-02-06 12:10:12,708 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93632.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:10:20,112 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2841, 1.8685, 2.7436, 2.2006, 2.5972, 2.2009, 1.7665, 1.0900], + device='cuda:0'), covar=tensor([0.4282, 0.4252, 0.1262, 0.2630, 0.1789, 0.2213, 0.1821, 0.4205], + device='cuda:0'), in_proj_covar=tensor([0.0893, 0.0872, 0.0725, 0.0851, 0.0935, 0.0802, 0.0700, 0.0767], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 12:10:21,470 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.6110, 2.2084, 3.2706, 2.6489, 3.0660, 2.4709, 1.8912, 1.7317], + device='cuda:0'), covar=tensor([0.3969, 0.4516, 0.1226, 0.2724, 0.1944, 0.2194, 0.1831, 0.4404], + device='cuda:0'), in_proj_covar=tensor([0.0893, 0.0871, 0.0725, 0.0851, 0.0934, 0.0802, 0.0699, 0.0767], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 12:10:26,597 INFO [optim.py:369] (0/4) 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 12:10:29,423 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93657.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:10:33,908 INFO [train.py:901] (0/4) Epoch 12, batch 4750, loss[loss=0.2122, simple_loss=0.2955, pruned_loss=0.06447, over 8135.00 frames. ], tot_loss[loss=0.2291, simple_loss=0.3055, pruned_loss=0.07636, over 1610311.84 frames. ], batch size: 22, lr: 6.36e-03, grad_scale: 8.0 +2023-02-06 12:10:34,135 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93664.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:10:51,753 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93689.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:11:00,473 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-06 12:11:02,149 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.62 vs. limit=2.0 +2023-02-06 12:11:02,490 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-06 12:11:04,759 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93708.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:11:09,471 INFO [train.py:901] (0/4) Epoch 12, batch 4800, loss[loss=0.2754, simple_loss=0.3514, pruned_loss=0.09966, over 8601.00 frames. ], tot_loss[loss=0.227, simple_loss=0.3038, pruned_loss=0.07514, over 1612202.32 frames. ], batch size: 49, lr: 6.35e-03, grad_scale: 8.0 +2023-02-06 12:11:19,234 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8985, 2.1356, 1.7742, 2.6158, 1.2037, 1.5696, 1.7496, 2.2707], + device='cuda:0'), covar=tensor([0.0723, 0.0772, 0.0966, 0.0382, 0.1081, 0.1221, 0.0871, 0.0625], + device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0210, 0.0253, 0.0214, 0.0215, 0.0249, 0.0255, 0.0217], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 12:11:22,630 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93733.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:11:28,709 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7249, 1.7415, 2.8181, 1.3046, 2.1271, 3.0514, 3.1283, 2.5352], + device='cuda:0'), covar=tensor([0.1040, 0.1213, 0.0362, 0.1961, 0.0830, 0.0313, 0.0583, 0.0710], + device='cuda:0'), in_proj_covar=tensor([0.0268, 0.0299, 0.0264, 0.0295, 0.0279, 0.0240, 0.0353, 0.0295], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 12:11:30,147 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93743.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:11:36,835 INFO [optim.py:369] (0/4) 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 12:11:44,130 INFO [train.py:901] (0/4) Epoch 12, batch 4850, loss[loss=0.265, simple_loss=0.3323, pruned_loss=0.0988, over 8247.00 frames. ], tot_loss[loss=0.2281, simple_loss=0.3048, pruned_loss=0.07575, over 1614189.59 frames. ], batch size: 22, lr: 6.35e-03, grad_scale: 8.0 +2023-02-06 12:11:47,091 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93768.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:11:47,784 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93769.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:11:53,077 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-06 12:12:00,551 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93788.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:12:04,716 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93794.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:12:18,636 INFO [train.py:901] (0/4) Epoch 12, batch 4900, loss[loss=0.2137, simple_loss=0.2971, pruned_loss=0.06514, over 8074.00 frames. ], tot_loss[loss=0.229, simple_loss=0.3057, pruned_loss=0.0762, over 1618311.87 frames. ], batch size: 21, lr: 6.35e-03, grad_scale: 8.0 +2023-02-06 12:12:42,681 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93848.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:12:45,724 INFO [optim.py:369] (0/4) 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 12:12:53,907 INFO [train.py:901] (0/4) Epoch 12, batch 4950, loss[loss=0.212, simple_loss=0.2868, pruned_loss=0.06857, over 8189.00 frames. ], tot_loss[loss=0.2291, simple_loss=0.3056, pruned_loss=0.07631, over 1620630.82 frames. ], batch size: 23, lr: 6.35e-03, grad_scale: 8.0 +2023-02-06 12:12:57,976 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3633, 2.8535, 2.4088, 4.0961, 1.9047, 1.9630, 2.4032, 3.1560], + device='cuda:0'), covar=tensor([0.0845, 0.0937, 0.0985, 0.0203, 0.1059, 0.1400, 0.1033, 0.0774], + device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0213, 0.0256, 0.0216, 0.0216, 0.0252, 0.0259, 0.0220], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 12:13:00,009 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93873.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:13:08,875 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.84 vs. limit=2.0 +2023-02-06 12:13:12,884 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 +2023-02-06 12:13:27,056 INFO [train.py:901] (0/4) Epoch 12, batch 5000, loss[loss=0.2032, simple_loss=0.2798, pruned_loss=0.06328, over 7977.00 frames. ], tot_loss[loss=0.2308, simple_loss=0.3065, pruned_loss=0.07752, over 1619739.75 frames. ], batch size: 21, lr: 6.35e-03, grad_scale: 8.0 +2023-02-06 12:13:55,395 INFO [optim.py:369] (0/4) 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 12:14:02,966 INFO [train.py:901] (0/4) Epoch 12, batch 5050, loss[loss=0.3399, simple_loss=0.3812, pruned_loss=0.1493, over 8365.00 frames. ], tot_loss[loss=0.2305, simple_loss=0.3069, pruned_loss=0.07709, over 1622035.44 frames. ], batch size: 24, lr: 6.35e-03, grad_scale: 8.0 +2023-02-06 12:14:27,668 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93999.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:14:28,288 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-94000.pt +2023-02-06 12:14:31,137 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-06 12:14:38,585 INFO [train.py:901] (0/4) Epoch 12, batch 5100, loss[loss=0.2616, simple_loss=0.3382, pruned_loss=0.09254, over 8315.00 frames. ], tot_loss[loss=0.2309, simple_loss=0.307, pruned_loss=0.07739, over 1617643.47 frames. ], batch size: 25, lr: 6.34e-03, grad_scale: 8.0 +2023-02-06 12:15:05,342 INFO [optim.py:369] (0/4) 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 12:15:13,507 INFO [train.py:901] (0/4) Epoch 12, batch 5150, loss[loss=0.2098, simple_loss=0.2885, pruned_loss=0.06555, over 7413.00 frames. ], tot_loss[loss=0.2307, simple_loss=0.3066, pruned_loss=0.0774, over 1617254.56 frames. ], batch size: 17, lr: 6.34e-03, grad_scale: 8.0 +2023-02-06 12:15:33,463 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 +2023-02-06 12:15:46,325 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4862, 2.3609, 4.1712, 1.2696, 2.9981, 1.9741, 1.6759, 2.7578], + device='cuda:0'), covar=tensor([0.1936, 0.2347, 0.0708, 0.4526, 0.1538, 0.3115, 0.2097, 0.2311], + device='cuda:0'), in_proj_covar=tensor([0.0492, 0.0526, 0.0535, 0.0587, 0.0618, 0.0558, 0.0478, 0.0617], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 12:15:47,487 INFO [train.py:901] (0/4) Epoch 12, batch 5200, loss[loss=0.226, simple_loss=0.3147, pruned_loss=0.06871, over 8454.00 frames. ], tot_loss[loss=0.2314, simple_loss=0.3072, pruned_loss=0.07777, over 1614406.41 frames. ], batch size: 27, lr: 6.34e-03, grad_scale: 8.0 +2023-02-06 12:15:58,650 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.5052, 2.9104, 2.4266, 4.0495, 1.8266, 2.1808, 2.2624, 3.2956], + device='cuda:0'), covar=tensor([0.0766, 0.0893, 0.0989, 0.0274, 0.1148, 0.1436, 0.1233, 0.0753], + device='cuda:0'), in_proj_covar=tensor([0.0237, 0.0216, 0.0258, 0.0219, 0.0220, 0.0256, 0.0263, 0.0224], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 12:15:59,923 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=94132.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:16:00,729 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7562, 1.6771, 2.6809, 1.3464, 2.1029, 2.8835, 2.8985, 2.4927], + device='cuda:0'), covar=tensor([0.0935, 0.1265, 0.0510, 0.1861, 0.1066, 0.0315, 0.0664, 0.0609], + device='cuda:0'), in_proj_covar=tensor([0.0267, 0.0296, 0.0262, 0.0291, 0.0275, 0.0238, 0.0349, 0.0292], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 12:16:14,668 INFO [optim.py:369] (0/4) 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 12:16:21,923 INFO [train.py:901] (0/4) Epoch 12, batch 5250, loss[loss=0.2611, simple_loss=0.3288, pruned_loss=0.09673, over 8329.00 frames. ], tot_loss[loss=0.232, simple_loss=0.3074, pruned_loss=0.07831, over 1615667.94 frames. ], batch size: 25, lr: 6.34e-03, grad_scale: 8.0 +2023-02-06 12:16:25,911 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 12, batch 5300, loss[loss=0.1894, simple_loss=0.2683, pruned_loss=0.05528, over 7532.00 frames. ], tot_loss[loss=0.2318, simple_loss=0.3073, pruned_loss=0.07816, over 1615096.71 frames. ], batch size: 18, lr: 6.34e-03, grad_scale: 8.0 +2023-02-06 12:17:13,243 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-06 12:17:15,480 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94241.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 12:17:19,500 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=94247.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:17:23,397 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 12, batch 5350, loss[loss=0.2326, simple_loss=0.3181, pruned_loss=0.07353, over 8792.00 frames. ], tot_loss[loss=0.2312, simple_loss=0.3073, pruned_loss=0.07758, over 1617170.96 frames. ], batch size: 30, lr: 6.34e-03, grad_scale: 8.0 +2023-02-06 12:17:42,437 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6401, 2.0240, 3.2058, 1.4194, 2.4528, 1.9940, 1.7190, 2.4292], + device='cuda:0'), covar=tensor([0.1582, 0.1949, 0.0636, 0.3513, 0.1458, 0.2519, 0.1674, 0.1900], + device='cuda:0'), in_proj_covar=tensor([0.0487, 0.0521, 0.0532, 0.0582, 0.0614, 0.0555, 0.0475, 0.0613], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 12:17:55,696 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2405, 1.6885, 4.4037, 1.7925, 2.5249, 4.9519, 5.0323, 4.3589], + device='cuda:0'), covar=tensor([0.1190, 0.1710, 0.0316, 0.2182, 0.1078, 0.0230, 0.0362, 0.0572], + device='cuda:0'), in_proj_covar=tensor([0.0266, 0.0298, 0.0261, 0.0292, 0.0274, 0.0236, 0.0348, 0.0291], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 12:18:05,202 INFO [train.py:901] (0/4) Epoch 12, batch 5400, loss[loss=0.2581, simple_loss=0.3356, pruned_loss=0.09031, over 8194.00 frames. ], tot_loss[loss=0.2319, simple_loss=0.3075, pruned_loss=0.07815, over 1618894.72 frames. ], batch size: 23, lr: 6.33e-03, grad_scale: 8.0 +2023-02-06 12:18:25,476 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=94343.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:18:32,239 INFO [optim.py:369] (0/4) 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 12:18:39,966 INFO [train.py:901] (0/4) Epoch 12, batch 5450, loss[loss=0.2225, simple_loss=0.3062, pruned_loss=0.06935, over 8094.00 frames. ], tot_loss[loss=0.231, simple_loss=0.3071, pruned_loss=0.07747, over 1614873.22 frames. ], batch size: 23, lr: 6.33e-03, grad_scale: 8.0 +2023-02-06 12:18:40,770 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2880, 2.1801, 1.7360, 1.9619, 1.8054, 1.4060, 1.7141, 1.6292], + device='cuda:0'), covar=tensor([0.1177, 0.0315, 0.0911, 0.0474, 0.0623, 0.1266, 0.0792, 0.0759], + device='cuda:0'), in_proj_covar=tensor([0.0344, 0.0232, 0.0311, 0.0293, 0.0296, 0.0320, 0.0334, 0.0301], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 12:18:41,415 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94366.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:19:12,392 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-06 12:19:15,756 INFO [train.py:901] (0/4) Epoch 12, batch 5500, loss[loss=0.2246, simple_loss=0.3113, pruned_loss=0.06896, over 8626.00 frames. ], tot_loss[loss=0.2294, simple_loss=0.3057, pruned_loss=0.07653, over 1613934.53 frames. ], batch size: 31, lr: 6.33e-03, grad_scale: 16.0 +2023-02-06 12:19:28,061 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3759, 1.6361, 4.1433, 2.0377, 2.4388, 4.7702, 4.7584, 4.1600], + device='cuda:0'), covar=tensor([0.0975, 0.1570, 0.0286, 0.1688, 0.1091, 0.0191, 0.0392, 0.0516], + device='cuda:0'), in_proj_covar=tensor([0.0267, 0.0297, 0.0260, 0.0291, 0.0274, 0.0235, 0.0347, 0.0290], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 12:19:43,155 INFO [optim.py:369] (0/4) 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 12:19:45,370 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=94458.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:19:46,246 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 +2023-02-06 12:19:49,122 INFO [train.py:901] (0/4) Epoch 12, batch 5550, loss[loss=0.2296, simple_loss=0.3027, pruned_loss=0.07824, over 7931.00 frames. ], tot_loss[loss=0.2304, simple_loss=0.3064, pruned_loss=0.07725, over 1610480.95 frames. ], batch size: 20, lr: 6.33e-03, grad_scale: 4.0 +2023-02-06 12:20:16,669 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=94503.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:20:17,399 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1466, 1.8452, 2.6145, 2.0556, 2.2844, 2.0448, 1.7078, 1.1684], + device='cuda:0'), covar=tensor([0.4162, 0.3880, 0.1320, 0.2835, 0.2220, 0.2498, 0.1849, 0.4064], + device='cuda:0'), in_proj_covar=tensor([0.0893, 0.0874, 0.0730, 0.0856, 0.0945, 0.0804, 0.0705, 0.0767], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 12:20:24,042 INFO [train.py:901] (0/4) Epoch 12, batch 5600, loss[loss=0.2235, simple_loss=0.3134, pruned_loss=0.06676, over 8465.00 frames. ], tot_loss[loss=0.2309, simple_loss=0.3072, pruned_loss=0.0773, over 1614241.86 frames. ], batch size: 25, lr: 6.33e-03, grad_scale: 8.0 +2023-02-06 12:20:35,223 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 12:21:00,666 INFO [train.py:901] (0/4) Epoch 12, batch 5650, loss[loss=0.2407, simple_loss=0.3122, pruned_loss=0.08459, over 8535.00 frames. ], tot_loss[loss=0.2311, simple_loss=0.3077, pruned_loss=0.07723, over 1620231.34 frames. ], batch size: 39, lr: 6.33e-03, grad_scale: 8.0 +2023-02-06 12:21:15,232 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=94585.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 12:21:21,326 WARNING [train.py:1067] (0/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-06 12:21:35,768 INFO [train.py:901] (0/4) Epoch 12, batch 5700, loss[loss=0.2332, simple_loss=0.3193, pruned_loss=0.07359, over 8468.00 frames. ], tot_loss[loss=0.2311, simple_loss=0.3075, pruned_loss=0.07736, over 1614552.14 frames. ], batch size: 27, lr: 6.32e-03, grad_scale: 8.0 +2023-02-06 12:22:04,729 INFO [optim.py:369] (0/4) 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 12:22:10,802 INFO [train.py:901] (0/4) Epoch 12, batch 5750, loss[loss=0.1856, simple_loss=0.2594, pruned_loss=0.0559, over 7648.00 frames. ], tot_loss[loss=0.2317, simple_loss=0.3081, pruned_loss=0.07769, over 1613343.91 frames. ], batch size: 19, lr: 6.32e-03, grad_scale: 8.0 +2023-02-06 12:22:25,646 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7499, 1.6163, 3.0011, 1.2576, 2.1173, 3.2919, 3.4219, 2.7514], + device='cuda:0'), covar=tensor([0.1226, 0.1560, 0.0395, 0.2221, 0.1054, 0.0296, 0.0460, 0.0730], + device='cuda:0'), in_proj_covar=tensor([0.0269, 0.0299, 0.0260, 0.0292, 0.0277, 0.0237, 0.0347, 0.0292], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 12:22:26,190 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-06 12:22:35,729 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=94700.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 12:22:42,475 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=94710.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:22:45,146 INFO [train.py:901] (0/4) Epoch 12, batch 5800, loss[loss=0.2376, simple_loss=0.3052, pruned_loss=0.08496, over 7404.00 frames. ], tot_loss[loss=0.2311, simple_loss=0.3077, pruned_loss=0.07726, over 1613457.03 frames. ], batch size: 17, lr: 6.32e-03, grad_scale: 8.0 +2023-02-06 12:22:45,331 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=94714.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:23:02,544 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=94739.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:23:13,643 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 12, batch 5850, loss[loss=0.2724, simple_loss=0.3382, pruned_loss=0.1033, over 8553.00 frames. ], tot_loss[loss=0.2331, simple_loss=0.309, pruned_loss=0.07855, over 1615017.88 frames. ], batch size: 31, lr: 6.32e-03, grad_scale: 8.0 +2023-02-06 12:23:54,273 INFO [train.py:901] (0/4) Epoch 12, batch 5900, loss[loss=0.221, simple_loss=0.3059, pruned_loss=0.06805, over 8241.00 frames. ], tot_loss[loss=0.2326, simple_loss=0.3088, pruned_loss=0.07821, over 1615541.46 frames. ], batch size: 24, lr: 6.32e-03, grad_scale: 8.0 +2023-02-06 12:24:01,747 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=94825.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:24:01,821 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0758, 1.8062, 2.4747, 2.0508, 2.3221, 2.0111, 1.6433, 0.9829], + device='cuda:0'), covar=tensor([0.4236, 0.3997, 0.1311, 0.2590, 0.1861, 0.2247, 0.1734, 0.4106], + device='cuda:0'), in_proj_covar=tensor([0.0892, 0.0880, 0.0732, 0.0856, 0.0939, 0.0803, 0.0706, 0.0769], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 12:24:08,300 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.4164, 1.1941, 4.5113, 1.6554, 3.9950, 3.7915, 4.0509, 3.9420], + device='cuda:0'), covar=tensor([0.0448, 0.4519, 0.0448, 0.3658, 0.1007, 0.0830, 0.0494, 0.0590], + device='cuda:0'), in_proj_covar=tensor([0.0513, 0.0581, 0.0596, 0.0544, 0.0623, 0.0535, 0.0525, 0.0584], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 12:24:22,270 INFO [optim.py:369] (0/4) 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 12:24:28,377 INFO [train.py:901] (0/4) Epoch 12, batch 5950, loss[loss=0.2533, simple_loss=0.3263, pruned_loss=0.09009, over 8455.00 frames. ], tot_loss[loss=0.2323, simple_loss=0.3082, pruned_loss=0.07818, over 1612519.39 frames. ], batch size: 27, lr: 6.32e-03, grad_scale: 8.0 +2023-02-06 12:25:03,798 INFO [train.py:901] (0/4) Epoch 12, batch 6000, loss[loss=0.2466, simple_loss=0.3288, pruned_loss=0.08226, over 8106.00 frames. ], tot_loss[loss=0.2314, simple_loss=0.3074, pruned_loss=0.07768, over 1615021.03 frames. ], batch size: 23, lr: 6.31e-03, grad_scale: 8.0 +2023-02-06 12:25:03,799 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 12:25:16,950 INFO [train.py:935] (0/4) Epoch 12, validation: loss=0.1862, simple_loss=0.286, pruned_loss=0.04318, over 944034.00 frames. +2023-02-06 12:25:16,952 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-06 12:25:44,732 INFO [optim.py:369] (0/4) 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 12:25:45,511 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=94956.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 12:25:50,751 INFO [train.py:901] (0/4) Epoch 12, batch 6050, loss[loss=0.2421, simple_loss=0.3238, pruned_loss=0.08024, over 8545.00 frames. ], tot_loss[loss=0.2313, simple_loss=0.3073, pruned_loss=0.07768, over 1615107.67 frames. ], batch size: 31, lr: 6.31e-03, grad_scale: 8.0 +2023-02-06 12:26:00,455 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.1247, 1.9749, 2.1093, 1.8097, 1.1939, 1.8076, 2.2712, 2.5497], + device='cuda:0'), covar=tensor([0.0430, 0.1102, 0.1533, 0.1289, 0.0587, 0.1410, 0.0624, 0.0500], + device='cuda:0'), in_proj_covar=tensor([0.0098, 0.0152, 0.0192, 0.0158, 0.0103, 0.0162, 0.0115, 0.0138], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 12:26:02,522 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=94981.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 12:26:25,550 INFO [train.py:901] (0/4) Epoch 12, batch 6100, loss[loss=0.2167, simple_loss=0.2986, pruned_loss=0.06739, over 7796.00 frames. ], tot_loss[loss=0.2315, simple_loss=0.3076, pruned_loss=0.07771, over 1616338.38 frames. ], batch size: 19, lr: 6.31e-03, grad_scale: 8.0 +2023-02-06 12:26:54,026 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-06 12:26:54,675 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 12, batch 6150, loss[loss=0.2239, simple_loss=0.3171, pruned_loss=0.06531, over 8084.00 frames. ], tot_loss[loss=0.2323, simple_loss=0.3079, pruned_loss=0.0783, over 1614470.52 frames. ], batch size: 21, lr: 6.31e-03, grad_scale: 8.0 +2023-02-06 12:27:12,217 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95081.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:27:29,632 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95106.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:27:34,683 INFO [train.py:901] (0/4) Epoch 12, batch 6200, loss[loss=0.2356, simple_loss=0.3207, pruned_loss=0.07527, over 8699.00 frames. ], tot_loss[loss=0.2325, simple_loss=0.308, pruned_loss=0.07846, over 1615588.51 frames. ], batch size: 34, lr: 6.31e-03, grad_scale: 8.0 +2023-02-06 12:27:41,099 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95123.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 12:27:46,636 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2031, 1.4757, 1.7304, 1.3522, 0.9709, 1.4504, 1.8466, 1.8713], + device='cuda:0'), covar=tensor([0.0467, 0.1212, 0.1726, 0.1371, 0.0601, 0.1456, 0.0615, 0.0552], + device='cuda:0'), in_proj_covar=tensor([0.0098, 0.0151, 0.0191, 0.0157, 0.0102, 0.0161, 0.0114, 0.0137], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 12:28:04,342 INFO [optim.py:369] (0/4) 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 12:28:07,870 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7700, 1.3820, 1.5199, 1.2720, 0.9044, 1.3832, 1.5360, 1.4738], + device='cuda:0'), covar=tensor([0.0525, 0.1223, 0.1792, 0.1470, 0.0623, 0.1569, 0.0677, 0.0612], + device='cuda:0'), in_proj_covar=tensor([0.0098, 0.0151, 0.0190, 0.0157, 0.0102, 0.0161, 0.0114, 0.0137], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 12:28:10,300 INFO [train.py:901] (0/4) Epoch 12, batch 6250, loss[loss=0.1975, simple_loss=0.2851, pruned_loss=0.0549, over 8458.00 frames. ], tot_loss[loss=0.2336, simple_loss=0.3088, pruned_loss=0.07918, over 1612867.68 frames. ], batch size: 25, lr: 6.31e-03, grad_scale: 8.0 +2023-02-06 12:28:18,457 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5509, 1.7205, 1.8208, 1.2502, 1.8960, 1.3120, 0.8500, 1.6076], + device='cuda:0'), covar=tensor([0.0338, 0.0223, 0.0136, 0.0316, 0.0223, 0.0481, 0.0504, 0.0204], + device='cuda:0'), in_proj_covar=tensor([0.0397, 0.0330, 0.0282, 0.0391, 0.0325, 0.0478, 0.0360, 0.0359], + device='cuda:0'), out_proj_covar=tensor([1.1204e-04, 9.0673e-05, 7.7960e-05, 1.0886e-04, 9.0804e-05, 1.4387e-04, + 1.0150e-04, 1.0065e-04], device='cuda:0') +2023-02-06 12:28:43,825 INFO [train.py:901] (0/4) Epoch 12, batch 6300, loss[loss=0.2387, simple_loss=0.3298, pruned_loss=0.07374, over 8450.00 frames. ], tot_loss[loss=0.2346, simple_loss=0.3096, pruned_loss=0.0798, over 1607661.41 frames. ], batch size: 27, lr: 6.30e-03, grad_scale: 8.0 +2023-02-06 12:29:13,414 INFO [optim.py:369] (0/4) 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 12:29:20,980 INFO [train.py:901] (0/4) Epoch 12, batch 6350, loss[loss=0.2123, simple_loss=0.3084, pruned_loss=0.05812, over 8496.00 frames. ], tot_loss[loss=0.2347, simple_loss=0.3103, pruned_loss=0.07956, over 1611355.98 frames. ], batch size: 26, lr: 6.30e-03, grad_scale: 8.0 +2023-02-06 12:29:30,713 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95278.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:29:36,272 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95286.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:29:55,396 INFO [train.py:901] (0/4) Epoch 12, batch 6400, loss[loss=0.2171, simple_loss=0.3005, pruned_loss=0.06682, over 8107.00 frames. ], tot_loss[loss=0.2306, simple_loss=0.3066, pruned_loss=0.07728, over 1611469.52 frames. ], batch size: 23, lr: 6.30e-03, grad_scale: 8.0 +2023-02-06 12:30:03,906 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.3399, 2.9588, 3.5183, 2.5758, 1.9717, 3.6444, 0.8238, 2.4342], + device='cuda:0'), covar=tensor([0.1640, 0.1178, 0.0420, 0.1946, 0.3538, 0.0270, 0.3280, 0.1574], + device='cuda:0'), in_proj_covar=tensor([0.0171, 0.0172, 0.0103, 0.0216, 0.0255, 0.0107, 0.0164, 0.0166], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 12:30:23,572 INFO [optim.py:369] (0/4) 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 12:30:25,092 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95357.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:30:29,660 INFO [train.py:901] (0/4) Epoch 12, batch 6450, loss[loss=0.206, simple_loss=0.2798, pruned_loss=0.06604, over 7653.00 frames. ], tot_loss[loss=0.2297, simple_loss=0.306, pruned_loss=0.07674, over 1610992.98 frames. ], batch size: 19, lr: 6.30e-03, grad_scale: 8.0 +2023-02-06 12:30:32,396 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9081, 1.4591, 3.2991, 1.2843, 2.3236, 3.5187, 3.7017, 3.0480], + device='cuda:0'), covar=tensor([0.1086, 0.1702, 0.0350, 0.2083, 0.0951, 0.0272, 0.0435, 0.0588], + device='cuda:0'), in_proj_covar=tensor([0.0265, 0.0296, 0.0257, 0.0288, 0.0271, 0.0235, 0.0345, 0.0286], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-06 12:31:05,025 INFO [train.py:901] (0/4) Epoch 12, batch 6500, loss[loss=0.306, simple_loss=0.3672, pruned_loss=0.1224, over 8610.00 frames. ], tot_loss[loss=0.2294, simple_loss=0.3057, pruned_loss=0.0766, over 1611975.29 frames. ], batch size: 39, lr: 6.30e-03, grad_scale: 8.0 +2023-02-06 12:31:18,807 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 +2023-02-06 12:31:22,443 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2396, 2.2282, 1.6639, 1.9526, 1.6978, 1.3201, 1.5211, 1.6482], + device='cuda:0'), covar=tensor([0.1157, 0.0331, 0.1063, 0.0494, 0.0672, 0.1381, 0.0912, 0.0758], + device='cuda:0'), in_proj_covar=tensor([0.0347, 0.0236, 0.0317, 0.0301, 0.0303, 0.0323, 0.0342, 0.0300], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 12:31:31,886 INFO [optim.py:369] (0/4) 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 12:31:37,945 INFO [train.py:901] (0/4) Epoch 12, batch 6550, loss[loss=0.2103, simple_loss=0.2962, pruned_loss=0.06218, over 8525.00 frames. ], tot_loss[loss=0.2306, simple_loss=0.3063, pruned_loss=0.07749, over 1606435.77 frames. ], batch size: 26, lr: 6.30e-03, grad_scale: 8.0 +2023-02-06 12:31:40,697 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=95467.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 12:32:04,346 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3377, 1.3355, 1.5310, 1.2830, 0.7089, 1.4202, 1.2941, 1.1651], + device='cuda:0'), covar=tensor([0.0530, 0.1212, 0.1671, 0.1363, 0.0565, 0.1432, 0.0622, 0.0631], + device='cuda:0'), in_proj_covar=tensor([0.0098, 0.0151, 0.0190, 0.0157, 0.0102, 0.0161, 0.0114, 0.0137], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 12:32:06,859 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-06 12:32:13,529 INFO [train.py:901] (0/4) Epoch 12, batch 6600, loss[loss=0.2435, simple_loss=0.3243, pruned_loss=0.08132, over 8291.00 frames. ], tot_loss[loss=0.2303, simple_loss=0.3066, pruned_loss=0.07704, over 1610519.09 frames. ], batch size: 23, lr: 6.29e-03, grad_scale: 8.0 +2023-02-06 12:32:25,783 WARNING [train.py:1067] (0/4) 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] (0/4) 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 12:32:46,211 INFO [train.py:901] (0/4) Epoch 12, batch 6650, loss[loss=0.252, simple_loss=0.3233, pruned_loss=0.09031, over 8465.00 frames. ], tot_loss[loss=0.2331, simple_loss=0.309, pruned_loss=0.07867, over 1607390.06 frames. ], batch size: 25, lr: 6.29e-03, grad_scale: 8.0 +2023-02-06 12:32:59,137 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95582.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 12:33:05,225 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.0437, 1.9256, 2.0122, 1.9994, 1.0555, 1.9259, 2.2080, 2.3090], + device='cuda:0'), covar=tensor([0.0392, 0.0991, 0.1520, 0.1108, 0.0539, 0.1232, 0.0590, 0.0490], + device='cuda:0'), in_proj_covar=tensor([0.0098, 0.0151, 0.0190, 0.0157, 0.0102, 0.0161, 0.0114, 0.0137], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 12:33:21,233 INFO [train.py:901] (0/4) Epoch 12, batch 6700, loss[loss=0.272, simple_loss=0.3249, pruned_loss=0.1096, over 7919.00 frames. ], tot_loss[loss=0.2324, simple_loss=0.3079, pruned_loss=0.07838, over 1607998.49 frames. ], batch size: 20, lr: 6.29e-03, grad_scale: 8.0 +2023-02-06 12:33:26,826 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=95622.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:33:33,696 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=95630.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:33:37,111 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95635.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:33:43,773 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9913, 2.2929, 1.9230, 2.8234, 1.3502, 1.6197, 1.9363, 2.4664], + device='cuda:0'), covar=tensor([0.0696, 0.0770, 0.0979, 0.0428, 0.1165, 0.1402, 0.1081, 0.0694], + device='cuda:0'), in_proj_covar=tensor([0.0236, 0.0214, 0.0255, 0.0218, 0.0215, 0.0255, 0.0261, 0.0221], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 12:33:50,488 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 12, batch 6750, loss[loss=0.1956, simple_loss=0.2635, pruned_loss=0.06388, over 7430.00 frames. ], tot_loss[loss=0.2306, simple_loss=0.3066, pruned_loss=0.07733, over 1607484.95 frames. ], batch size: 17, lr: 6.29e-03, grad_scale: 8.0 +2023-02-06 12:34:22,171 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=95701.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:34:24,378 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1947, 1.8013, 2.5068, 2.0324, 2.2993, 2.1043, 1.7349, 1.1142], + device='cuda:0'), covar=tensor([0.4153, 0.4026, 0.1319, 0.2794, 0.1835, 0.2203, 0.1762, 0.4047], + device='cuda:0'), in_proj_covar=tensor([0.0894, 0.0883, 0.0736, 0.0867, 0.0941, 0.0811, 0.0708, 0.0775], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 12:34:30,620 INFO [train.py:901] (0/4) Epoch 12, batch 6800, loss[loss=0.2279, simple_loss=0.3046, pruned_loss=0.07563, over 7971.00 frames. ], tot_loss[loss=0.231, simple_loss=0.3072, pruned_loss=0.07743, over 1609896.40 frames. ], batch size: 21, lr: 6.29e-03, grad_scale: 8.0 +2023-02-06 12:34:40,704 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-06 12:34:47,118 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95737.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:34:53,229 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 12:35:06,296 INFO [train.py:901] (0/4) Epoch 12, batch 6850, loss[loss=0.2447, simple_loss=0.3052, pruned_loss=0.09211, over 7548.00 frames. ], tot_loss[loss=0.2301, simple_loss=0.3063, pruned_loss=0.07696, over 1609702.22 frames. ], batch size: 18, lr: 6.29e-03, grad_scale: 8.0 +2023-02-06 12:35:26,874 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-06 12:35:32,367 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2399, 1.2922, 1.4924, 1.2185, 0.7222, 1.3104, 1.1257, 1.1254], + device='cuda:0'), covar=tensor([0.0547, 0.1221, 0.1714, 0.1391, 0.0584, 0.1490, 0.0693, 0.0634], + device='cuda:0'), in_proj_covar=tensor([0.0099, 0.0152, 0.0191, 0.0159, 0.0103, 0.0162, 0.0115, 0.0138], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0008, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 12:35:40,360 INFO [train.py:901] (0/4) Epoch 12, batch 6900, loss[loss=0.2074, simple_loss=0.2824, pruned_loss=0.06627, over 7820.00 frames. ], tot_loss[loss=0.2307, simple_loss=0.3071, pruned_loss=0.07721, over 1612655.77 frames. ], batch size: 20, lr: 6.29e-03, grad_scale: 8.0 +2023-02-06 12:35:41,928 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95816.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:35:51,215 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95830.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 12:35:57,273 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95838.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 12:36:03,155 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2993, 1.9255, 2.6803, 2.1581, 2.4902, 2.1841, 1.8052, 1.2266], + device='cuda:0'), covar=tensor([0.3730, 0.3732, 0.1216, 0.2706, 0.1827, 0.2112, 0.1622, 0.4228], + device='cuda:0'), in_proj_covar=tensor([0.0887, 0.0878, 0.0731, 0.0858, 0.0935, 0.0806, 0.0704, 0.0770], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 12:36:08,028 INFO [optim.py:369] (0/4) 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 12:36:14,243 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95863.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 12:36:14,702 INFO [train.py:901] (0/4) Epoch 12, batch 6950, loss[loss=0.2134, simple_loss=0.2994, pruned_loss=0.06371, over 7981.00 frames. ], tot_loss[loss=0.2311, simple_loss=0.3075, pruned_loss=0.0774, over 1611755.39 frames. ], batch size: 21, lr: 6.28e-03, grad_scale: 8.0 +2023-02-06 12:36:15,461 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95865.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:36:34,591 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-06 12:36:47,613 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0471, 2.7151, 3.1155, 1.3211, 3.2185, 1.7297, 1.5415, 2.1183], + device='cuda:0'), covar=tensor([0.0696, 0.0245, 0.0278, 0.0668, 0.0310, 0.0818, 0.0654, 0.0414], + device='cuda:0'), in_proj_covar=tensor([0.0398, 0.0332, 0.0284, 0.0394, 0.0326, 0.0487, 0.0363, 0.0365], + device='cuda:0'), out_proj_covar=tensor([1.1241e-04, 9.1129e-05, 7.8523e-05, 1.0925e-04, 9.1131e-05, 1.4664e-04, + 1.0241e-04, 1.0220e-04], device='cuda:0') +2023-02-06 12:36:48,726 INFO [train.py:901] (0/4) Epoch 12, batch 7000, loss[loss=0.2697, simple_loss=0.3436, pruned_loss=0.09792, over 8202.00 frames. ], tot_loss[loss=0.2323, simple_loss=0.3084, pruned_loss=0.07814, over 1614175.84 frames. ], batch size: 23, lr: 6.28e-03, grad_scale: 8.0 +2023-02-06 12:37:17,422 INFO [optim.py:369] (0/4) 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 12:37:23,279 INFO [train.py:901] (0/4) Epoch 12, batch 7050, loss[loss=0.2283, simple_loss=0.3138, pruned_loss=0.07147, over 8454.00 frames. ], tot_loss[loss=0.2314, simple_loss=0.3076, pruned_loss=0.07761, over 1615113.77 frames. ], batch size: 27, lr: 6.28e-03, grad_scale: 8.0 +2023-02-06 12:37:24,149 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8710, 1.7533, 2.5101, 1.5264, 1.1646, 2.5026, 0.3860, 1.3636], + device='cuda:0'), covar=tensor([0.2446, 0.1953, 0.0391, 0.2547, 0.4945, 0.0436, 0.3668, 0.2163], + device='cuda:0'), in_proj_covar=tensor([0.0172, 0.0176, 0.0105, 0.0219, 0.0258, 0.0110, 0.0167, 0.0168], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 12:37:34,000 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=95979.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:37:44,053 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95993.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:37:46,010 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8036, 1.4544, 3.3502, 1.3031, 2.3546, 3.7529, 3.7420, 3.2048], + device='cuda:0'), covar=tensor([0.1138, 0.1645, 0.0371, 0.2197, 0.1000, 0.0223, 0.0505, 0.0549], + device='cuda:0'), in_proj_covar=tensor([0.0268, 0.0297, 0.0261, 0.0290, 0.0273, 0.0235, 0.0347, 0.0286], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-06 12:37:47,385 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95998.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:37:48,733 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-96000.pt +2023-02-06 12:37:50,531 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96001.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:37:58,499 INFO [train.py:901] (0/4) Epoch 12, batch 7100, loss[loss=0.216, simple_loss=0.2999, pruned_loss=0.06603, over 8287.00 frames. ], tot_loss[loss=0.2313, simple_loss=0.3076, pruned_loss=0.07753, over 1615345.87 frames. ], batch size: 23, lr: 6.28e-03, grad_scale: 8.0 +2023-02-06 12:38:01,399 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96018.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:38:06,974 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96026.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:38:26,967 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 12, batch 7150, loss[loss=0.2127, simple_loss=0.2813, pruned_loss=0.07207, over 7973.00 frames. ], tot_loss[loss=0.2306, simple_loss=0.3071, pruned_loss=0.0771, over 1614352.94 frames. ], batch size: 21, lr: 6.28e-03, grad_scale: 8.0 +2023-02-06 12:38:38,700 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96072.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:38:42,266 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 +2023-02-06 12:38:54,774 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96094.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:38:56,849 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96097.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:38:58,158 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96098.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:39:08,658 INFO [train.py:901] (0/4) Epoch 12, batch 7200, loss[loss=0.2036, simple_loss=0.2666, pruned_loss=0.07032, over 7444.00 frames. ], tot_loss[loss=0.2294, simple_loss=0.3061, pruned_loss=0.07633, over 1616706.18 frames. ], batch size: 17, lr: 6.28e-03, grad_scale: 8.0 +2023-02-06 12:39:10,832 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4736, 1.9610, 3.2892, 1.2528, 2.4621, 1.9673, 1.5512, 2.3599], + device='cuda:0'), covar=tensor([0.1849, 0.2333, 0.0806, 0.4188, 0.1767, 0.2960, 0.2040, 0.2415], + device='cuda:0'), in_proj_covar=tensor([0.0489, 0.0528, 0.0536, 0.0579, 0.0621, 0.0554, 0.0476, 0.0613], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 12:39:36,232 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 12, batch 7250, loss[loss=0.2104, simple_loss=0.2972, pruned_loss=0.06178, over 8307.00 frames. ], tot_loss[loss=0.2282, simple_loss=0.305, pruned_loss=0.07572, over 1612717.54 frames. ], batch size: 25, lr: 6.27e-03, grad_scale: 8.0 +2023-02-06 12:39:49,509 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=96174.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 12:39:55,424 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96183.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:40:06,506 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 +2023-02-06 12:40:14,127 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=96209.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:40:15,131 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.10 vs. limit=5.0 +2023-02-06 12:40:17,290 INFO [train.py:901] (0/4) Epoch 12, batch 7300, loss[loss=0.2501, simple_loss=0.3204, pruned_loss=0.08985, over 7541.00 frames. ], tot_loss[loss=0.229, simple_loss=0.3048, pruned_loss=0.07653, over 1606941.04 frames. ], batch size: 18, lr: 6.27e-03, grad_scale: 8.0 +2023-02-06 12:40:45,402 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 12, batch 7350, loss[loss=0.2142, simple_loss=0.2975, pruned_loss=0.06539, over 7963.00 frames. ], tot_loss[loss=0.2285, simple_loss=0.3045, pruned_loss=0.0763, over 1604190.37 frames. ], batch size: 21, lr: 6.27e-03, grad_scale: 8.0 +2023-02-06 12:41:09,292 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96289.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 12:41:15,847 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-06 12:41:26,501 INFO [train.py:901] (0/4) Epoch 12, batch 7400, loss[loss=0.2604, simple_loss=0.3424, pruned_loss=0.08914, over 8597.00 frames. ], tot_loss[loss=0.2298, simple_loss=0.3061, pruned_loss=0.07673, over 1610052.26 frames. ], batch size: 39, lr: 6.27e-03, grad_scale: 8.0 +2023-02-06 12:41:33,416 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96324.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:41:36,511 WARNING [train.py:1067] (0/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-06 12:41:47,016 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=96342.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:41:47,086 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96342.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:41:52,319 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96350.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:41:55,369 INFO [optim.py:369] (0/4) 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 12:42:01,561 INFO [train.py:901] (0/4) Epoch 12, batch 7450, loss[loss=0.2521, simple_loss=0.3261, pruned_loss=0.0891, over 8510.00 frames. ], tot_loss[loss=0.2308, simple_loss=0.3066, pruned_loss=0.07753, over 1608229.21 frames. ], batch size: 26, lr: 6.27e-03, grad_scale: 8.0 +2023-02-06 12:42:09,242 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96375.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:42:15,384 WARNING [train.py:1067] (0/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-06 12:42:31,406 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96408.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:42:35,240 INFO [train.py:901] (0/4) Epoch 12, batch 7500, loss[loss=0.2253, simple_loss=0.306, pruned_loss=0.07232, over 8372.00 frames. ], tot_loss[loss=0.2315, simple_loss=0.307, pruned_loss=0.07798, over 1612066.26 frames. ], batch size: 24, lr: 6.27e-03, grad_scale: 8.0 +2023-02-06 12:42:47,518 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 +2023-02-06 12:42:54,771 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=96442.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:43:03,962 INFO [optim.py:369] (0/4) 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 12:43:05,492 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96457.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:43:09,793 INFO [train.py:901] (0/4) Epoch 12, batch 7550, loss[loss=0.2755, simple_loss=0.3432, pruned_loss=0.1039, over 8464.00 frames. ], tot_loss[loss=0.2325, simple_loss=0.3077, pruned_loss=0.07871, over 1614045.45 frames. ], batch size: 25, lr: 6.26e-03, grad_scale: 16.0 +2023-02-06 12:43:42,979 INFO [train.py:901] (0/4) Epoch 12, batch 7600, loss[loss=0.2157, simple_loss=0.2876, pruned_loss=0.07191, over 7659.00 frames. ], tot_loss[loss=0.2324, simple_loss=0.3075, pruned_loss=0.07862, over 1613041.17 frames. ], batch size: 19, lr: 6.26e-03, grad_scale: 16.0 +2023-02-06 12:43:52,495 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=96527.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:44:05,562 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96545.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 12:44:11,790 INFO [optim.py:369] (0/4) 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 12:44:13,324 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96557.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:44:18,532 INFO [train.py:901] (0/4) Epoch 12, batch 7650, loss[loss=0.2428, simple_loss=0.3269, pruned_loss=0.07941, over 8713.00 frames. ], tot_loss[loss=0.2323, simple_loss=0.3076, pruned_loss=0.0785, over 1610381.04 frames. ], batch size: 34, lr: 6.26e-03, grad_scale: 16.0 +2023-02-06 12:44:23,312 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96570.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 12:44:29,833 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96580.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:44:47,079 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96605.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:44:53,017 INFO [train.py:901] (0/4) Epoch 12, batch 7700, loss[loss=0.2336, simple_loss=0.3076, pruned_loss=0.0798, over 8289.00 frames. ], tot_loss[loss=0.2312, simple_loss=0.3066, pruned_loss=0.0779, over 1607992.06 frames. ], batch size: 23, lr: 6.26e-03, grad_scale: 16.0 +2023-02-06 12:45:12,723 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96642.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:45:13,563 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-06 12:45:21,231 INFO [optim.py:369] (0/4) 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 12:45:23,902 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-06 12:45:27,882 INFO [train.py:901] (0/4) Epoch 12, batch 7750, loss[loss=0.2429, simple_loss=0.3207, pruned_loss=0.0825, over 8472.00 frames. ], tot_loss[loss=0.2302, simple_loss=0.3059, pruned_loss=0.07728, over 1609977.10 frames. ], batch size: 29, lr: 6.26e-03, grad_scale: 16.0 +2023-02-06 12:45:42,849 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=96686.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:45:57,501 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 +2023-02-06 12:46:02,106 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96713.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:46:02,590 INFO [train.py:901] (0/4) Epoch 12, batch 7800, loss[loss=0.2405, simple_loss=0.3257, pruned_loss=0.07767, over 8471.00 frames. ], tot_loss[loss=0.23, simple_loss=0.3059, pruned_loss=0.077, over 1611227.49 frames. ], batch size: 25, lr: 6.26e-03, grad_scale: 16.0 +2023-02-06 12:46:19,274 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96738.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:46:28,444 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=96752.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:46:30,316 INFO [optim.py:369] (0/4) 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 12:46:36,572 INFO [train.py:901] (0/4) Epoch 12, batch 7850, loss[loss=0.1846, simple_loss=0.2607, pruned_loss=0.05429, over 7312.00 frames. ], tot_loss[loss=0.2301, simple_loss=0.3059, pruned_loss=0.07715, over 1606812.11 frames. ], batch size: 16, lr: 6.25e-03, grad_scale: 16.0 +2023-02-06 12:46:46,557 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 +2023-02-06 12:46:55,800 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.53 vs. limit=5.0 +2023-02-06 12:46:56,877 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96794.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:47:01,916 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96801.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:47:09,552 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 +2023-02-06 12:47:10,025 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96813.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:47:10,502 INFO [train.py:901] (0/4) Epoch 12, batch 7900, loss[loss=0.2401, simple_loss=0.3212, pruned_loss=0.07947, over 8464.00 frames. ], tot_loss[loss=0.231, simple_loss=0.3073, pruned_loss=0.07736, over 1607930.21 frames. ], batch size: 29, lr: 6.25e-03, grad_scale: 16.0 +2023-02-06 12:47:27,502 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 12:47:44,799 INFO [train.py:901] (0/4) Epoch 12, batch 7950, loss[loss=0.2297, simple_loss=0.306, pruned_loss=0.07668, over 8032.00 frames. ], tot_loss[loss=0.2318, simple_loss=0.308, pruned_loss=0.07786, over 1611514.04 frames. ], batch size: 22, lr: 6.25e-03, grad_scale: 16.0 +2023-02-06 12:47:47,027 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96867.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:48:07,408 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96898.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:48:15,446 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.61 vs. limit=2.0 +2023-02-06 12:48:17,695 INFO [train.py:901] (0/4) Epoch 12, batch 8000, loss[loss=0.2166, simple_loss=0.305, pruned_loss=0.06408, over 8498.00 frames. ], tot_loss[loss=0.2316, simple_loss=0.3077, pruned_loss=0.07774, over 1610595.96 frames. ], batch size: 28, lr: 6.25e-03, grad_scale: 16.0 +2023-02-06 12:48:23,660 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 12:48:51,267 INFO [train.py:901] (0/4) Epoch 12, batch 8050, loss[loss=0.2649, simple_loss=0.3341, pruned_loss=0.09788, over 6474.00 frames. ], tot_loss[loss=0.2322, simple_loss=0.3078, pruned_loss=0.07829, over 1603047.43 frames. ], batch size: 71, lr: 6.25e-03, grad_scale: 16.0 +2023-02-06 12:49:15,168 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/epoch-12.pt +2023-02-06 12:49:25,986 WARNING [train.py:1067] (0/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-06 12:49:29,795 INFO [train.py:901] (0/4) Epoch 13, batch 0, loss[loss=0.1982, simple_loss=0.286, pruned_loss=0.05518, over 8193.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.286, pruned_loss=0.05518, over 8193.00 frames. ], batch size: 23, lr: 6.00e-03, grad_scale: 16.0 +2023-02-06 12:49:29,796 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 12:49:40,736 INFO [train.py:935] (0/4) Epoch 13, validation: loss=0.1867, simple_loss=0.2865, pruned_loss=0.04345, over 944034.00 frames. +2023-02-06 12:49:40,737 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-06 12:49:48,848 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 +2023-02-06 12:49:55,394 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-06 12:49:55,528 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97018.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:49:57,150 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 +2023-02-06 12:50:14,919 INFO [train.py:901] (0/4) Epoch 13, batch 50, loss[loss=0.2767, simple_loss=0.3425, pruned_loss=0.1055, over 8193.00 frames. ], tot_loss[loss=0.2331, simple_loss=0.3098, pruned_loss=0.07822, over 362754.68 frames. ], batch size: 23, lr: 6.00e-03, grad_scale: 16.0 +2023-02-06 12:50:20,331 INFO [optim.py:369] (0/4) 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 12:50:21,933 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97057.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:50:29,214 WARNING [train.py:1067] (0/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-06 12:50:41,099 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 13, batch 100, loss[loss=0.1948, simple_loss=0.2757, pruned_loss=0.05702, over 8084.00 frames. ], tot_loss[loss=0.2319, simple_loss=0.309, pruned_loss=0.07741, over 641805.87 frames. ], batch size: 21, lr: 6.00e-03, grad_scale: 16.0 +2023-02-06 12:50:52,984 WARNING [train.py:1067] (0/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-06 12:51:02,199 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.1823, 4.1925, 3.7374, 1.8210, 3.6792, 3.6520, 3.8426, 3.3754], + device='cuda:0'), covar=tensor([0.0818, 0.0556, 0.1079, 0.4552, 0.1032, 0.1030, 0.1189, 0.0930], + device='cuda:0'), in_proj_covar=tensor([0.0475, 0.0387, 0.0397, 0.0496, 0.0390, 0.0392, 0.0385, 0.0337], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 12:51:09,044 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97123.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:51:18,985 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 13, batch 150, loss[loss=0.2316, simple_loss=0.2972, pruned_loss=0.08302, over 7156.00 frames. ], tot_loss[loss=0.2295, simple_loss=0.3063, pruned_loss=0.07638, over 853778.08 frames. ], batch size: 16, lr: 6.00e-03, grad_scale: 16.0 +2023-02-06 12:51:25,604 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97148.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:51:30,113 INFO [optim.py:369] (0/4) 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 12:51:58,451 INFO [train.py:901] (0/4) Epoch 13, batch 200, loss[loss=0.2444, simple_loss=0.3226, pruned_loss=0.08311, over 8503.00 frames. ], tot_loss[loss=0.2292, simple_loss=0.3063, pruned_loss=0.07606, over 1028403.26 frames. ], batch size: 26, lr: 6.00e-03, grad_scale: 16.0 +2023-02-06 12:52:09,192 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3716, 1.3389, 2.3138, 1.2978, 2.0891, 2.4341, 2.5462, 2.1244], + device='cuda:0'), covar=tensor([0.1001, 0.1213, 0.0463, 0.1916, 0.0727, 0.0393, 0.0656, 0.0703], + device='cuda:0'), in_proj_covar=tensor([0.0269, 0.0301, 0.0263, 0.0292, 0.0275, 0.0239, 0.0357, 0.0291], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-06 12:52:33,408 INFO [train.py:901] (0/4) Epoch 13, batch 250, loss[loss=0.2249, simple_loss=0.3045, pruned_loss=0.07269, over 8482.00 frames. ], tot_loss[loss=0.229, simple_loss=0.3058, pruned_loss=0.07608, over 1161562.23 frames. ], batch size: 49, lr: 6.00e-03, grad_scale: 16.0 +2023-02-06 12:52:37,623 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 12:52:46,026 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-06 12:52:53,399 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5964, 1.8207, 1.9240, 1.3148, 2.0462, 1.4486, 0.7302, 1.7488], + device='cuda:0'), covar=tensor([0.0420, 0.0227, 0.0171, 0.0350, 0.0247, 0.0523, 0.0535, 0.0203], + device='cuda:0'), in_proj_covar=tensor([0.0397, 0.0334, 0.0283, 0.0390, 0.0319, 0.0476, 0.0355, 0.0360], + device='cuda:0'), out_proj_covar=tensor([1.1174e-04, 9.1578e-05, 7.8097e-05, 1.0806e-04, 8.8825e-05, 1.4247e-04, + 1.0009e-04, 1.0050e-04], device='cuda:0') +2023-02-06 12:52:54,012 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97278.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:52:54,538 WARNING [train.py:1067] (0/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-06 12:53:06,364 INFO [train.py:901] (0/4) Epoch 13, batch 300, loss[loss=0.1854, simple_loss=0.2742, pruned_loss=0.04835, over 7811.00 frames. ], tot_loss[loss=0.2296, simple_loss=0.3062, pruned_loss=0.07649, over 1265654.02 frames. ], batch size: 20, lr: 5.99e-03, grad_scale: 16.0 +2023-02-06 12:53:15,189 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2858, 2.1646, 1.7015, 2.0109, 1.7079, 1.4035, 1.6100, 1.7011], + device='cuda:0'), covar=tensor([0.1127, 0.0347, 0.0984, 0.0462, 0.0635, 0.1239, 0.0853, 0.0742], + device='cuda:0'), in_proj_covar=tensor([0.0346, 0.0236, 0.0321, 0.0301, 0.0304, 0.0322, 0.0342, 0.0302], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 12:53:41,539 INFO [train.py:901] (0/4) Epoch 13, batch 350, loss[loss=0.1978, simple_loss=0.2799, pruned_loss=0.05781, over 8102.00 frames. ], tot_loss[loss=0.2294, simple_loss=0.3062, pruned_loss=0.07624, over 1342925.12 frames. ], batch size: 23, lr: 5.99e-03, grad_scale: 16.0 +2023-02-06 12:53:46,924 INFO [optim.py:369] (0/4) 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 12:53:50,448 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97360.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:53:51,744 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=97362.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:53:53,793 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97365.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:54:11,873 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 13, batch 400, loss[loss=0.2054, simple_loss=0.2892, pruned_loss=0.06083, over 7649.00 frames. ], tot_loss[loss=0.2294, simple_loss=0.3067, pruned_loss=0.07607, over 1405078.81 frames. ], batch size: 19, lr: 5.99e-03, grad_scale: 16.0 +2023-02-06 12:54:51,190 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.9212, 1.4956, 5.9457, 2.0729, 5.3329, 5.0387, 5.5219, 5.3517], + device='cuda:0'), covar=tensor([0.0387, 0.4622, 0.0356, 0.3406, 0.0933, 0.0817, 0.0438, 0.0444], + device='cuda:0'), in_proj_covar=tensor([0.0507, 0.0588, 0.0595, 0.0545, 0.0625, 0.0531, 0.0523, 0.0584], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 12:54:51,725 INFO [train.py:901] (0/4) Epoch 13, batch 450, loss[loss=0.2249, simple_loss=0.3117, pruned_loss=0.06899, over 8096.00 frames. ], tot_loss[loss=0.2293, simple_loss=0.3064, pruned_loss=0.07609, over 1455347.88 frames. ], batch size: 23, lr: 5.99e-03, grad_scale: 16.0 +2023-02-06 12:54:57,098 INFO [optim.py:369] (0/4) 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 12:55:04,802 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7626, 1.5083, 2.7953, 1.2977, 2.1692, 2.9727, 3.1327, 2.5166], + device='cuda:0'), covar=tensor([0.1019, 0.1429, 0.0410, 0.2070, 0.0879, 0.0339, 0.0558, 0.0712], + device='cuda:0'), in_proj_covar=tensor([0.0269, 0.0301, 0.0266, 0.0294, 0.0275, 0.0240, 0.0359, 0.0291], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 12:55:04,850 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2626, 2.1311, 1.6307, 1.9163, 1.7811, 1.2644, 1.6274, 1.6426], + device='cuda:0'), covar=tensor([0.1136, 0.0344, 0.1074, 0.0487, 0.0581, 0.1348, 0.0878, 0.0737], + device='cuda:0'), in_proj_covar=tensor([0.0346, 0.0236, 0.0319, 0.0301, 0.0302, 0.0323, 0.0341, 0.0302], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 12:55:06,273 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8241, 1.8816, 2.2682, 1.7516, 1.2711, 2.3097, 0.3914, 1.4190], + device='cuda:0'), covar=tensor([0.2336, 0.1259, 0.0457, 0.1633, 0.3921, 0.0396, 0.2988, 0.1721], + device='cuda:0'), in_proj_covar=tensor([0.0172, 0.0172, 0.0103, 0.0219, 0.0258, 0.0109, 0.0164, 0.0169], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 12:55:13,252 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97477.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:55:26,756 INFO [train.py:901] (0/4) Epoch 13, batch 500, loss[loss=0.2155, simple_loss=0.2945, pruned_loss=0.06824, over 8356.00 frames. ], tot_loss[loss=0.228, simple_loss=0.3055, pruned_loss=0.0752, over 1492857.38 frames. ], batch size: 24, lr: 5.99e-03, grad_scale: 16.0 +2023-02-06 12:55:35,480 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97509.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:55:52,950 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97534.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:55:53,115 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=2.05 vs. limit=2.0 +2023-02-06 12:56:01,766 INFO [train.py:901] (0/4) Epoch 13, batch 550, loss[loss=0.2726, simple_loss=0.3475, pruned_loss=0.09878, over 8192.00 frames. ], tot_loss[loss=0.2285, simple_loss=0.3053, pruned_loss=0.07584, over 1517976.45 frames. ], batch size: 23, lr: 5.99e-03, grad_scale: 16.0 +2023-02-06 12:56:07,713 INFO [optim.py:369] (0/4) 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 12:56:20,093 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97573.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:56:36,765 INFO [train.py:901] (0/4) Epoch 13, batch 600, loss[loss=0.2384, simple_loss=0.3238, pruned_loss=0.07646, over 8200.00 frames. ], tot_loss[loss=0.2274, simple_loss=0.3048, pruned_loss=0.07498, over 1537765.47 frames. ], batch size: 23, lr: 5.98e-03, grad_scale: 16.0 +2023-02-06 12:56:53,684 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=97622.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:56:55,690 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-06 12:57:10,256 INFO [train.py:901] (0/4) Epoch 13, batch 650, loss[loss=0.2125, simple_loss=0.2962, pruned_loss=0.06438, over 8208.00 frames. ], tot_loss[loss=0.2284, simple_loss=0.3053, pruned_loss=0.07573, over 1553017.72 frames. ], batch size: 23, lr: 5.98e-03, grad_scale: 16.0 +2023-02-06 12:57:16,274 INFO [optim.py:369] (0/4) 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 12:57:25,803 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3039, 1.9534, 1.4310, 1.8087, 1.5237, 1.1767, 1.4562, 1.6634], + device='cuda:0'), covar=tensor([0.1192, 0.0421, 0.1250, 0.0538, 0.0746, 0.1591, 0.1040, 0.0781], + device='cuda:0'), in_proj_covar=tensor([0.0348, 0.0237, 0.0322, 0.0302, 0.0304, 0.0325, 0.0344, 0.0304], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 12:57:42,538 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97692.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 12:57:45,736 INFO [train.py:901] (0/4) Epoch 13, batch 700, loss[loss=0.1796, simple_loss=0.2499, pruned_loss=0.05465, over 7935.00 frames. ], tot_loss[loss=0.2281, simple_loss=0.3051, pruned_loss=0.07554, over 1569119.57 frames. ], batch size: 20, lr: 5.98e-03, grad_scale: 16.0 +2023-02-06 12:57:51,261 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=97704.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:57:54,504 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=97709.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:58:10,695 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97733.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:58:12,603 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=97736.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:58:13,385 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97737.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:58:19,707 INFO [train.py:901] (0/4) Epoch 13, batch 750, loss[loss=0.2335, simple_loss=0.3179, pruned_loss=0.0746, over 8643.00 frames. ], tot_loss[loss=0.2289, simple_loss=0.3062, pruned_loss=0.07584, over 1584547.17 frames. ], batch size: 34, lr: 5.98e-03, grad_scale: 16.0 +2023-02-06 12:58:25,045 INFO [optim.py:369] (0/4) 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] (0/4) 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] (0/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-06 12:58:49,007 WARNING [train.py:1067] (0/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-06 12:58:54,230 INFO [train.py:901] (0/4) Epoch 13, batch 800, loss[loss=0.2694, simple_loss=0.3387, pruned_loss=0.1001, over 8341.00 frames. ], tot_loss[loss=0.2275, simple_loss=0.305, pruned_loss=0.07497, over 1594444.06 frames. ], batch size: 26, lr: 5.98e-03, grad_scale: 16.0 +2023-02-06 12:58:56,585 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.37 vs. limit=5.0 +2023-02-06 12:59:10,089 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97819.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:59:14,045 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97824.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:59:29,482 INFO [train.py:901] (0/4) Epoch 13, batch 850, loss[loss=0.2065, simple_loss=0.2919, pruned_loss=0.06057, over 8043.00 frames. ], tot_loss[loss=0.2273, simple_loss=0.305, pruned_loss=0.07484, over 1600910.94 frames. ], batch size: 22, lr: 5.98e-03, grad_scale: 16.0 +2023-02-06 12:59:32,328 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97851.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 12:59:35,503 INFO [optim.py:369] (0/4) 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 13:00:03,797 INFO [train.py:901] (0/4) Epoch 13, batch 900, loss[loss=0.2479, simple_loss=0.3175, pruned_loss=0.08914, over 8500.00 frames. ], tot_loss[loss=0.2272, simple_loss=0.3046, pruned_loss=0.07492, over 1600671.68 frames. ], batch size: 28, lr: 5.98e-03, grad_scale: 8.0 +2023-02-06 13:00:09,699 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97906.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:00:18,213 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=97917.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:00:39,215 INFO [train.py:901] (0/4) Epoch 13, batch 950, loss[loss=0.1983, simple_loss=0.2912, pruned_loss=0.05271, over 8278.00 frames. ], tot_loss[loss=0.227, simple_loss=0.3045, pruned_loss=0.07476, over 1606068.70 frames. ], batch size: 23, lr: 5.97e-03, grad_scale: 8.0 +2023-02-06 13:00:45,304 INFO [optim.py:369] (0/4) 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 13:00:59,997 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.1696, 4.1183, 3.7590, 1.7608, 3.6800, 3.7131, 3.7509, 3.3275], + device='cuda:0'), covar=tensor([0.0817, 0.0558, 0.1060, 0.4850, 0.0948, 0.0867, 0.1278, 0.0990], + device='cuda:0'), in_proj_covar=tensor([0.0474, 0.0387, 0.0394, 0.0489, 0.0389, 0.0392, 0.0385, 0.0341], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 13:01:08,766 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-06 13:01:11,112 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97993.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:01:13,716 INFO [train.py:901] (0/4) Epoch 13, batch 1000, loss[loss=0.2128, simple_loss=0.3024, pruned_loss=0.06161, over 8473.00 frames. ], tot_loss[loss=0.227, simple_loss=0.3048, pruned_loss=0.07458, over 1606270.84 frames. ], batch size: 25, lr: 5.97e-03, grad_scale: 8.0 +2023-02-06 13:01:15,919 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-98000.pt +2023-02-06 13:01:29,780 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98018.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:01:39,955 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98032.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:01:43,175 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=98036.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 13:01:44,381 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-06 13:01:50,447 INFO [train.py:901] (0/4) Epoch 13, batch 1050, loss[loss=0.2795, simple_loss=0.3404, pruned_loss=0.1093, over 7125.00 frames. ], tot_loss[loss=0.2265, simple_loss=0.304, pruned_loss=0.07451, over 1605292.27 frames. ], batch size: 71, lr: 5.97e-03, grad_scale: 8.0 +2023-02-06 13:01:56,534 INFO [optim.py:369] (0/4) 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 13:01:57,235 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-06 13:02:10,161 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98075.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:02:13,450 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98080.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:02:24,721 INFO [train.py:901] (0/4) Epoch 13, batch 1100, loss[loss=0.2403, simple_loss=0.3261, pruned_loss=0.07729, over 8106.00 frames. ], tot_loss[loss=0.2252, simple_loss=0.3027, pruned_loss=0.07384, over 1606904.71 frames. ], batch size: 23, lr: 5.97e-03, grad_scale: 8.0 +2023-02-06 13:02:26,939 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98105.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:02:31,641 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98107.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:02:42,954 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98124.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:02:49,125 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98132.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:02:59,477 INFO [train.py:901] (0/4) Epoch 13, batch 1150, loss[loss=0.2118, simple_loss=0.2744, pruned_loss=0.07465, over 7410.00 frames. ], tot_loss[loss=0.2262, simple_loss=0.3037, pruned_loss=0.07435, over 1609741.76 frames. ], batch size: 17, lr: 5.97e-03, grad_scale: 8.0 +2023-02-06 13:03:02,365 INFO [zipformer.py:1185] (0/4) 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] (0/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-06 13:03:06,090 INFO [optim.py:369] (0/4) 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 13:03:14,927 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0827, 1.2448, 1.1938, 0.6139, 1.2386, 1.0196, 0.0737, 1.1814], + device='cuda:0'), covar=tensor([0.0306, 0.0269, 0.0256, 0.0430, 0.0327, 0.0705, 0.0569, 0.0230], + device='cuda:0'), in_proj_covar=tensor([0.0402, 0.0337, 0.0289, 0.0397, 0.0325, 0.0484, 0.0360, 0.0362], + device='cuda:0'), 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:0') +2023-02-06 13:03:34,170 INFO [train.py:901] (0/4) Epoch 13, batch 1200, loss[loss=0.1897, simple_loss=0.2589, pruned_loss=0.06026, over 7409.00 frames. ], tot_loss[loss=0.2253, simple_loss=0.3024, pruned_loss=0.07408, over 1606332.76 frames. ], batch size: 17, lr: 5.97e-03, grad_scale: 8.0 +2023-02-06 13:03:46,026 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 +2023-02-06 13:04:06,227 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.5203, 1.1629, 4.6726, 1.6481, 4.1003, 3.9128, 4.2107, 4.0476], + device='cuda:0'), covar=tensor([0.0478, 0.4902, 0.0401, 0.3775, 0.1019, 0.0848, 0.0523, 0.0573], + device='cuda:0'), in_proj_covar=tensor([0.0510, 0.0587, 0.0598, 0.0544, 0.0618, 0.0531, 0.0522, 0.0579], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 13:04:08,152 INFO [train.py:901] (0/4) Epoch 13, batch 1250, loss[loss=0.2514, simple_loss=0.3377, pruned_loss=0.08252, over 8459.00 frames. ], tot_loss[loss=0.2271, simple_loss=0.304, pruned_loss=0.07511, over 1612664.13 frames. ], batch size: 27, lr: 5.96e-03, grad_scale: 8.0 +2023-02-06 13:04:10,201 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=98250.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:04:14,080 INFO [optim.py:369] (0/4) 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 13:04:36,935 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 13, batch 1300, loss[loss=0.2935, simple_loss=0.3547, pruned_loss=0.1162, over 8451.00 frames. ], tot_loss[loss=0.227, simple_loss=0.3038, pruned_loss=0.07506, over 1612592.20 frames. ], batch size: 27, lr: 5.96e-03, grad_scale: 8.0 +2023-02-06 13:04:44,912 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2999, 1.8710, 2.8775, 2.2569, 2.4934, 2.1120, 1.7428, 1.1716], + device='cuda:0'), covar=tensor([0.4373, 0.4651, 0.1211, 0.2926, 0.2159, 0.2761, 0.2008, 0.4674], + device='cuda:0'), in_proj_covar=tensor([0.0893, 0.0883, 0.0741, 0.0861, 0.0939, 0.0810, 0.0706, 0.0771], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 13:04:54,305 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0573, 1.5447, 1.5482, 1.3963, 0.9077, 1.3656, 1.7164, 1.6596], + device='cuda:0'), covar=tensor([0.0466, 0.1152, 0.1736, 0.1337, 0.0586, 0.1482, 0.0670, 0.0573], + device='cuda:0'), in_proj_covar=tensor([0.0100, 0.0153, 0.0193, 0.0158, 0.0102, 0.0164, 0.0116, 0.0139], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 13:04:54,335 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98313.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:05:12,874 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3365, 2.5215, 1.8345, 2.0648, 1.9827, 1.4733, 1.7148, 1.9707], + device='cuda:0'), covar=tensor([0.1275, 0.0325, 0.0968, 0.0572, 0.0767, 0.1380, 0.1008, 0.0766], + device='cuda:0'), in_proj_covar=tensor([0.0338, 0.0231, 0.0313, 0.0295, 0.0296, 0.0317, 0.0334, 0.0298], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 13:05:16,645 INFO [train.py:901] (0/4) Epoch 13, batch 1350, loss[loss=0.1894, simple_loss=0.2647, pruned_loss=0.05708, over 7212.00 frames. ], tot_loss[loss=0.2269, simple_loss=0.304, pruned_loss=0.07491, over 1614621.30 frames. ], batch size: 16, lr: 5.96e-03, grad_scale: 8.0 +2023-02-06 13:05:23,232 INFO [optim.py:369] (0/4) 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 13:05:29,565 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 13, batch 1400, loss[loss=0.2691, simple_loss=0.34, pruned_loss=0.09914, over 8283.00 frames. ], tot_loss[loss=0.2271, simple_loss=0.3041, pruned_loss=0.07501, over 1615343.70 frames. ], batch size: 23, lr: 5.96e-03, grad_scale: 8.0 +2023-02-06 13:05:59,393 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98407.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 13:06:14,063 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98428.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:06:16,845 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98432.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 13:06:26,899 INFO [train.py:901] (0/4) Epoch 13, batch 1450, loss[loss=0.2513, simple_loss=0.3301, pruned_loss=0.08622, over 8179.00 frames. ], tot_loss[loss=0.2267, simple_loss=0.3039, pruned_loss=0.07474, over 1611329.68 frames. ], batch size: 23, lr: 5.96e-03, grad_scale: 8.0 +2023-02-06 13:06:32,939 INFO [optim.py:369] (0/4) 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] (0/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-06 13:06:41,214 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=98468.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:07:02,148 INFO [train.py:901] (0/4) Epoch 13, batch 1500, loss[loss=0.2097, simple_loss=0.291, pruned_loss=0.06426, over 8519.00 frames. ], tot_loss[loss=0.2254, simple_loss=0.3025, pruned_loss=0.07415, over 1610048.53 frames. ], batch size: 28, lr: 5.96e-03, grad_scale: 8.0 +2023-02-06 13:07:28,243 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4897, 1.9401, 3.3815, 1.3137, 2.5393, 1.8862, 1.5861, 2.2719], + device='cuda:0'), covar=tensor([0.1759, 0.2242, 0.0730, 0.4119, 0.1529, 0.3017, 0.1991, 0.2170], + device='cuda:0'), in_proj_covar=tensor([0.0492, 0.0532, 0.0538, 0.0590, 0.0626, 0.0561, 0.0483, 0.0613], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 13:07:36,058 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.6819, 4.7127, 4.1899, 2.1584, 4.0271, 4.1854, 4.3431, 3.9378], + device='cuda:0'), covar=tensor([0.0611, 0.0515, 0.0885, 0.4228, 0.0907, 0.0897, 0.1110, 0.0744], + device='cuda:0'), in_proj_covar=tensor([0.0466, 0.0382, 0.0391, 0.0483, 0.0383, 0.0388, 0.0380, 0.0337], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 13:07:37,386 INFO [train.py:901] (0/4) Epoch 13, batch 1550, loss[loss=0.2159, simple_loss=0.2849, pruned_loss=0.07339, over 8090.00 frames. ], tot_loss[loss=0.2265, simple_loss=0.3033, pruned_loss=0.07492, over 1606635.73 frames. ], batch size: 21, lr: 5.96e-03, grad_scale: 8.0 +2023-02-06 13:07:43,379 INFO [optim.py:369] (0/4) 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 13:07:46,948 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98561.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:07:47,839 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.43 vs. limit=5.0 +2023-02-06 13:08:02,048 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98583.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:08:11,760 INFO [train.py:901] (0/4) Epoch 13, batch 1600, loss[loss=0.1992, simple_loss=0.282, pruned_loss=0.0582, over 7971.00 frames. ], tot_loss[loss=0.2269, simple_loss=0.3035, pruned_loss=0.07511, over 1606227.03 frames. ], batch size: 21, lr: 5.95e-03, grad_scale: 8.0 +2023-02-06 13:08:27,711 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.1850, 1.1009, 1.2608, 1.1025, 0.9151, 1.3219, 0.0582, 0.8731], + device='cuda:0'), covar=tensor([0.2158, 0.1807, 0.0628, 0.1154, 0.3652, 0.0590, 0.2939, 0.1533], + device='cuda:0'), in_proj_covar=tensor([0.0172, 0.0175, 0.0104, 0.0221, 0.0262, 0.0112, 0.0164, 0.0168], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 13:08:29,731 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98621.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:08:46,790 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98646.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:08:47,291 INFO [train.py:901] (0/4) Epoch 13, batch 1650, loss[loss=0.1894, simple_loss=0.2758, pruned_loss=0.05145, over 8196.00 frames. ], tot_loss[loss=0.2293, simple_loss=0.3057, pruned_loss=0.07643, over 1610801.58 frames. ], batch size: 23, lr: 5.95e-03, grad_scale: 8.0 +2023-02-06 13:08:53,405 INFO [optim.py:369] (0/4) 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 13:09:07,750 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.76 vs. limit=5.0 +2023-02-06 13:09:20,790 INFO [train.py:901] (0/4) Epoch 13, batch 1700, loss[loss=0.1854, simple_loss=0.2707, pruned_loss=0.04999, over 8198.00 frames. ], tot_loss[loss=0.2296, simple_loss=0.3063, pruned_loss=0.07648, over 1613285.85 frames. ], batch size: 23, lr: 5.95e-03, grad_scale: 8.0 +2023-02-06 13:09:49,366 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6668, 1.6270, 2.0288, 1.5368, 1.2491, 2.0357, 0.3356, 1.2629], + device='cuda:0'), covar=tensor([0.3027, 0.1710, 0.0547, 0.1648, 0.3600, 0.0540, 0.2898, 0.1675], + device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0175, 0.0104, 0.0220, 0.0260, 0.0112, 0.0163, 0.0167], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 13:09:57,386 INFO [train.py:901] (0/4) Epoch 13, batch 1750, loss[loss=0.2293, simple_loss=0.3097, pruned_loss=0.07445, over 8498.00 frames. ], tot_loss[loss=0.2292, simple_loss=0.3057, pruned_loss=0.07636, over 1613019.01 frames. ], batch size: 26, lr: 5.95e-03, grad_scale: 8.0 +2023-02-06 13:10:03,433 INFO [optim.py:369] (0/4) 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 13:10:15,119 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=98772.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:10:27,277 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.8031, 1.2321, 5.9177, 2.1968, 5.2242, 5.0087, 5.4782, 5.3288], + device='cuda:0'), covar=tensor([0.0487, 0.5155, 0.0332, 0.3358, 0.1019, 0.0848, 0.0460, 0.0467], + device='cuda:0'), in_proj_covar=tensor([0.0510, 0.0584, 0.0597, 0.0550, 0.0620, 0.0532, 0.0522, 0.0585], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 13:10:31,881 INFO [train.py:901] (0/4) Epoch 13, batch 1800, loss[loss=0.1886, simple_loss=0.2671, pruned_loss=0.05505, over 7981.00 frames. ], tot_loss[loss=0.2282, simple_loss=0.3052, pruned_loss=0.07558, over 1614144.81 frames. ], batch size: 21, lr: 5.95e-03, grad_scale: 8.0 +2023-02-06 13:11:01,397 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 13, batch 1850, loss[loss=0.2694, simple_loss=0.3268, pruned_loss=0.106, over 7532.00 frames. ], tot_loss[loss=0.2278, simple_loss=0.3046, pruned_loss=0.07556, over 1612029.55 frames. ], batch size: 18, lr: 5.95e-03, grad_scale: 8.0 +2023-02-06 13:11:13,496 INFO [optim.py:369] (0/4) 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 13:11:18,919 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98864.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:11:34,567 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98887.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:11:41,187 INFO [train.py:901] (0/4) Epoch 13, batch 1900, loss[loss=0.2908, simple_loss=0.358, pruned_loss=0.1118, over 7184.00 frames. ], tot_loss[loss=0.228, simple_loss=0.3047, pruned_loss=0.07563, over 1613865.67 frames. ], batch size: 71, lr: 5.95e-03, grad_scale: 8.0 +2023-02-06 13:11:46,600 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98908.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:11:51,815 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98913.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:12:11,194 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 13, batch 1950, loss[loss=0.2292, simple_loss=0.3137, pruned_loss=0.07232, over 8607.00 frames. ], tot_loss[loss=0.2277, simple_loss=0.3048, pruned_loss=0.07526, over 1616536.06 frames. ], batch size: 31, lr: 5.94e-03, grad_scale: 8.0 +2023-02-06 13:12:21,304 INFO [optim.py:369] (0/4) 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 13:12:23,981 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-06 13:12:44,583 WARNING [train.py:1067] (0/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-06 13:12:45,384 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98989.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:12:50,400 INFO [train.py:901] (0/4) Epoch 13, batch 2000, loss[loss=0.2607, simple_loss=0.3409, pruned_loss=0.09026, over 8187.00 frames. ], tot_loss[loss=0.2284, simple_loss=0.3055, pruned_loss=0.07565, over 1613384.75 frames. ], batch size: 23, lr: 5.94e-03, grad_scale: 8.0 +2023-02-06 13:13:06,435 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99020.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:13:24,243 INFO [train.py:901] (0/4) Epoch 13, batch 2050, loss[loss=0.2117, simple_loss=0.2918, pruned_loss=0.06576, over 8301.00 frames. ], tot_loss[loss=0.2254, simple_loss=0.3027, pruned_loss=0.07408, over 1607754.05 frames. ], batch size: 23, lr: 5.94e-03, grad_scale: 8.0 +2023-02-06 13:13:28,300 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8116, 1.9860, 2.0488, 1.2211, 2.2208, 1.5535, 0.5430, 1.9276], + device='cuda:0'), covar=tensor([0.0351, 0.0221, 0.0191, 0.0432, 0.0226, 0.0622, 0.0609, 0.0175], + device='cuda:0'), in_proj_covar=tensor([0.0393, 0.0330, 0.0286, 0.0394, 0.0321, 0.0478, 0.0355, 0.0359], + device='cuda:0'), out_proj_covar=tensor([1.1037e-04, 8.9948e-05, 7.8544e-05, 1.0883e-04, 8.9356e-05, 1.4268e-04, + 9.9842e-05, 1.0003e-04], device='cuda:0') +2023-02-06 13:13:28,455 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 +2023-02-06 13:13:30,114 INFO [optim.py:369] (0/4) 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 13:13:39,432 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99069.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:13:48,845 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.0762, 1.9969, 2.0920, 1.8406, 1.1032, 1.9244, 2.2757, 2.3245], + device='cuda:0'), covar=tensor([0.0383, 0.1077, 0.1451, 0.1184, 0.0548, 0.1296, 0.0592, 0.0480], + device='cuda:0'), in_proj_covar=tensor([0.0100, 0.0153, 0.0192, 0.0158, 0.0102, 0.0163, 0.0115, 0.0139], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0008, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 13:13:58,777 INFO [train.py:901] (0/4) Epoch 13, batch 2100, loss[loss=0.3112, simple_loss=0.358, pruned_loss=0.1322, over 8520.00 frames. ], tot_loss[loss=0.2243, simple_loss=0.3014, pruned_loss=0.07363, over 1604121.85 frames. ], batch size: 28, lr: 5.94e-03, grad_scale: 8.0 +2023-02-06 13:13:59,949 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 +2023-02-06 13:14:31,269 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99143.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:14:33,890 INFO [train.py:901] (0/4) Epoch 13, batch 2150, loss[loss=0.231, simple_loss=0.3168, pruned_loss=0.07259, over 8195.00 frames. ], tot_loss[loss=0.2273, simple_loss=0.3036, pruned_loss=0.07543, over 1601227.75 frames. ], batch size: 23, lr: 5.94e-03, grad_scale: 8.0 +2023-02-06 13:14:39,911 INFO [optim.py:369] (0/4) 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 13:14:48,055 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99168.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:15:07,844 INFO [train.py:901] (0/4) Epoch 13, batch 2200, loss[loss=0.2172, simple_loss=0.2954, pruned_loss=0.06952, over 8240.00 frames. ], tot_loss[loss=0.2258, simple_loss=0.3023, pruned_loss=0.07464, over 1599204.57 frames. ], batch size: 22, lr: 5.94e-03, grad_scale: 8.0 +2023-02-06 13:15:39,425 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 +2023-02-06 13:15:43,034 INFO [train.py:901] (0/4) Epoch 13, batch 2250, loss[loss=0.2705, simple_loss=0.3494, pruned_loss=0.09579, over 8309.00 frames. ], tot_loss[loss=0.2268, simple_loss=0.3033, pruned_loss=0.07514, over 1599008.53 frames. ], batch size: 26, lr: 5.93e-03, grad_scale: 8.0 +2023-02-06 13:15:43,152 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99247.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:15:46,442 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 13:15:49,661 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=99257.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:16:02,579 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99276.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:16:09,205 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99286.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:16:16,491 INFO [train.py:901] (0/4) Epoch 13, batch 2300, loss[loss=0.2151, simple_loss=0.2812, pruned_loss=0.07449, over 7686.00 frames. ], tot_loss[loss=0.2262, simple_loss=0.3029, pruned_loss=0.07471, over 1602936.76 frames. ], batch size: 18, lr: 5.93e-03, grad_scale: 8.0 +2023-02-06 13:16:19,363 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99301.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:16:43,123 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=99333.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:16:49,159 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99342.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:16:52,410 INFO [train.py:901] (0/4) Epoch 13, batch 2350, loss[loss=0.2714, simple_loss=0.338, pruned_loss=0.1024, over 8670.00 frames. ], tot_loss[loss=0.226, simple_loss=0.3028, pruned_loss=0.07464, over 1606334.13 frames. ], batch size: 49, lr: 5.93e-03, grad_scale: 8.0 +2023-02-06 13:16:58,385 INFO [optim.py:369] (0/4) 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 13:17:06,751 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99367.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:17:10,156 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99372.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:17:26,590 INFO [train.py:901] (0/4) Epoch 13, batch 2400, loss[loss=0.2414, simple_loss=0.3218, pruned_loss=0.0805, over 8020.00 frames. ], tot_loss[loss=0.2266, simple_loss=0.3031, pruned_loss=0.07506, over 1606910.60 frames. ], batch size: 22, lr: 5.93e-03, grad_scale: 8.0 +2023-02-06 13:17:26,696 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.1800, 4.1494, 3.7832, 1.8217, 3.7620, 3.7185, 3.8006, 3.5783], + device='cuda:0'), covar=tensor([0.0733, 0.0623, 0.0987, 0.4833, 0.0781, 0.0915, 0.1363, 0.0735], + device='cuda:0'), in_proj_covar=tensor([0.0479, 0.0393, 0.0397, 0.0497, 0.0391, 0.0394, 0.0384, 0.0343], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 13:17:37,524 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=99413.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:17:55,975 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99438.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:18:01,834 INFO [train.py:901] (0/4) Epoch 13, batch 2450, loss[loss=0.1913, simple_loss=0.2591, pruned_loss=0.0617, over 7702.00 frames. ], tot_loss[loss=0.2273, simple_loss=0.3041, pruned_loss=0.07528, over 1612521.40 frames. ], batch size: 18, lr: 5.93e-03, grad_scale: 8.0 +2023-02-06 13:18:02,705 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 13:18:29,570 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99486.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 13:18:36,514 INFO [train.py:901] (0/4) Epoch 13, batch 2500, loss[loss=0.2169, simple_loss=0.2847, pruned_loss=0.07456, over 7522.00 frames. ], tot_loss[loss=0.2252, simple_loss=0.302, pruned_loss=0.07417, over 1611652.55 frames. ], batch size: 18, lr: 5.93e-03, grad_scale: 8.0 +2023-02-06 13:18:57,692 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99528.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:19:10,257 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0844, 1.2544, 1.4485, 1.2098, 1.0332, 1.3122, 1.7718, 1.6993], + device='cuda:0'), covar=tensor([0.0571, 0.1807, 0.2538, 0.1890, 0.0722, 0.2182, 0.0791, 0.0685], + device='cuda:0'), in_proj_covar=tensor([0.0099, 0.0153, 0.0192, 0.0158, 0.0102, 0.0164, 0.0115, 0.0139], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 13:19:10,748 INFO [train.py:901] (0/4) Epoch 13, batch 2550, loss[loss=0.2807, simple_loss=0.3544, pruned_loss=0.1035, over 8604.00 frames. ], tot_loss[loss=0.2264, simple_loss=0.3033, pruned_loss=0.07475, over 1615716.43 frames. ], batch size: 39, lr: 5.93e-03, grad_scale: 8.0 +2023-02-06 13:19:17,199 INFO [optim.py:369] (0/4) 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 13:19:37,743 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99586.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:19:41,075 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 13, batch 2600, loss[loss=0.3143, simple_loss=0.3746, pruned_loss=0.127, over 7403.00 frames. ], tot_loss[loss=0.2272, simple_loss=0.3039, pruned_loss=0.07523, over 1613755.59 frames. ], batch size: 71, lr: 5.92e-03, grad_scale: 8.0 +2023-02-06 13:19:47,528 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 +2023-02-06 13:19:52,603 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99607.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:20:03,513 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99623.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:20:06,919 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=99630.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:20:19,560 INFO [train.py:901] (0/4) Epoch 13, batch 2650, loss[loss=0.2041, simple_loss=0.2955, pruned_loss=0.05636, over 8236.00 frames. ], tot_loss[loss=0.2271, simple_loss=0.3038, pruned_loss=0.07518, over 1613523.48 frames. ], batch size: 22, lr: 5.92e-03, grad_scale: 8.0 +2023-02-06 13:20:20,457 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99648.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:20:23,639 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99653.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:20:25,437 INFO [optim.py:369] (0/4) 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 13:20:46,670 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=99686.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:20:54,457 INFO [train.py:901] (0/4) Epoch 13, batch 2700, loss[loss=0.1871, simple_loss=0.2716, pruned_loss=0.05128, over 7803.00 frames. ], tot_loss[loss=0.2272, simple_loss=0.3038, pruned_loss=0.07527, over 1611398.63 frames. ], batch size: 19, lr: 5.92e-03, grad_scale: 8.0 +2023-02-06 13:20:55,208 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99698.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:20:59,198 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99704.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:21:00,470 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99706.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:21:16,801 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99729.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:21:27,701 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99745.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:21:28,870 INFO [train.py:901] (0/4) Epoch 13, batch 2750, loss[loss=0.252, simple_loss=0.3257, pruned_loss=0.08912, over 8366.00 frames. ], tot_loss[loss=0.2267, simple_loss=0.3033, pruned_loss=0.07507, over 1610387.80 frames. ], batch size: 24, lr: 5.92e-03, grad_scale: 8.0 +2023-02-06 13:21:34,777 INFO [optim.py:369] (0/4) 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 13:21:49,874 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5475, 1.8063, 2.8204, 1.3534, 2.0568, 1.8530, 1.5774, 1.8887], + device='cuda:0'), covar=tensor([0.1636, 0.2057, 0.0626, 0.3717, 0.1485, 0.2739, 0.1849, 0.1928], + device='cuda:0'), in_proj_covar=tensor([0.0489, 0.0530, 0.0539, 0.0588, 0.0622, 0.0559, 0.0481, 0.0612], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 13:21:52,457 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=99782.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:21:53,979 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99784.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:22:03,191 INFO [train.py:901] (0/4) Epoch 13, batch 2800, loss[loss=0.2084, simple_loss=0.2956, pruned_loss=0.06062, over 8336.00 frames. ], tot_loss[loss=0.2268, simple_loss=0.3034, pruned_loss=0.07509, over 1610808.27 frames. ], batch size: 26, lr: 5.92e-03, grad_scale: 8.0 +2023-02-06 13:22:06,156 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99801.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:22:11,456 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99809.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:22:26,099 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=99830.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 13:22:37,605 INFO [train.py:901] (0/4) Epoch 13, batch 2850, loss[loss=0.2377, simple_loss=0.3146, pruned_loss=0.08037, over 8133.00 frames. ], tot_loss[loss=0.2281, simple_loss=0.3048, pruned_loss=0.07566, over 1609657.82 frames. ], batch size: 22, lr: 5.92e-03, grad_scale: 8.0 +2023-02-06 13:22:43,877 INFO [optim.py:369] (0/4) 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 13:23:11,496 INFO [train.py:901] (0/4) Epoch 13, batch 2900, loss[loss=0.2204, simple_loss=0.3089, pruned_loss=0.06592, over 8364.00 frames. ], tot_loss[loss=0.2283, simple_loss=0.3052, pruned_loss=0.07568, over 1611466.47 frames. ], batch size: 24, lr: 5.92e-03, grad_scale: 16.0 +2023-02-06 13:23:11,675 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=99930.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:23:45,006 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99945.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 13:23:46,151 INFO [train.py:901] (0/4) Epoch 13, batch 2950, loss[loss=0.1965, simple_loss=0.2872, pruned_loss=0.05295, over 8581.00 frames. ], tot_loss[loss=0.2287, simple_loss=0.3054, pruned_loss=0.07604, over 1610385.98 frames. ], batch size: 31, lr: 5.91e-03, grad_scale: 16.0 +2023-02-06 13:23:48,976 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=99951.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:23:52,242 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-06 13:23:52,908 INFO [optim.py:369] (0/4) 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 13:23:57,273 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99962.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:24:14,467 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99987.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:24:21,168 INFO [train.py:901] (0/4) Epoch 13, batch 3000, loss[loss=0.1908, simple_loss=0.2642, pruned_loss=0.05872, over 7639.00 frames. ], tot_loss[loss=0.2273, simple_loss=0.3042, pruned_loss=0.07519, over 1613120.20 frames. ], batch size: 19, lr: 5.91e-03, grad_scale: 16.0 +2023-02-06 13:24:21,169 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 13:24:33,568 INFO [train.py:935] (0/4) Epoch 13, validation: loss=0.1841, simple_loss=0.2841, pruned_loss=0.04204, over 944034.00 frames. +2023-02-06 13:24:33,569 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-06 13:24:35,838 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-100000.pt +2023-02-06 13:24:37,720 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100001.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:24:53,947 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100025.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 13:24:55,300 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100026.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:25:05,689 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=100042.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:25:07,819 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100045.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:25:09,026 INFO [train.py:901] (0/4) Epoch 13, batch 3050, loss[loss=0.2053, simple_loss=0.2896, pruned_loss=0.06054, over 7929.00 frames. ], tot_loss[loss=0.2272, simple_loss=0.3045, pruned_loss=0.07495, over 1616361.46 frames. ], batch size: 20, lr: 5.91e-03, grad_scale: 16.0 +2023-02-06 13:25:15,852 INFO [optim.py:369] (0/4) 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 13:25:16,802 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100057.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:25:22,889 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100066.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:25:29,663 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.2184, 4.1975, 3.7284, 1.6740, 3.7200, 3.7697, 3.7454, 3.4603], + device='cuda:0'), covar=tensor([0.0748, 0.0571, 0.1096, 0.4897, 0.0817, 0.1067, 0.1314, 0.0883], + device='cuda:0'), in_proj_covar=tensor([0.0481, 0.0393, 0.0397, 0.0497, 0.0390, 0.0396, 0.0388, 0.0346], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 13:25:31,829 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1372, 2.9209, 3.2810, 1.3134, 3.3788, 2.0124, 1.4318, 2.1128], + device='cuda:0'), covar=tensor([0.0578, 0.0232, 0.0167, 0.0572, 0.0258, 0.0592, 0.0699, 0.0400], + device='cuda:0'), in_proj_covar=tensor([0.0393, 0.0335, 0.0287, 0.0394, 0.0323, 0.0481, 0.0358, 0.0362], + device='cuda:0'), out_proj_covar=tensor([1.1003e-04, 9.1589e-05, 7.8762e-05, 1.0881e-04, 8.9588e-05, 1.4336e-04, + 1.0074e-04, 1.0061e-04], device='cuda:0') +2023-02-06 13:25:34,605 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100082.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:25:40,508 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 +2023-02-06 13:25:44,735 INFO [train.py:901] (0/4) Epoch 13, batch 3100, loss[loss=0.2109, simple_loss=0.2994, pruned_loss=0.06124, over 8293.00 frames. ], tot_loss[loss=0.2266, simple_loss=0.3046, pruned_loss=0.0743, over 1617989.43 frames. ], batch size: 23, lr: 5.91e-03, grad_scale: 16.0 +2023-02-06 13:26:19,853 INFO [train.py:901] (0/4) Epoch 13, batch 3150, loss[loss=0.2243, simple_loss=0.2969, pruned_loss=0.07586, over 8084.00 frames. ], tot_loss[loss=0.2274, simple_loss=0.3049, pruned_loss=0.07494, over 1616690.91 frames. ], batch size: 21, lr: 5.91e-03, grad_scale: 16.0 +2023-02-06 13:26:24,137 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100153.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:26:25,935 INFO [optim.py:369] (0/4) 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 13:26:26,821 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100157.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:26:41,732 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100178.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:26:55,156 INFO [train.py:901] (0/4) Epoch 13, batch 3200, loss[loss=0.212, simple_loss=0.3008, pruned_loss=0.06162, over 8567.00 frames. ], tot_loss[loss=0.2282, simple_loss=0.3054, pruned_loss=0.07549, over 1617485.41 frames. ], batch size: 31, lr: 5.91e-03, grad_scale: 16.0 +2023-02-06 13:26:58,245 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100201.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 13:27:15,351 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100226.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 13:27:29,407 INFO [train.py:901] (0/4) Epoch 13, batch 3250, loss[loss=0.2679, simple_loss=0.3426, pruned_loss=0.0966, over 8717.00 frames. ], tot_loss[loss=0.2291, simple_loss=0.3065, pruned_loss=0.07585, over 1617920.88 frames. ], batch size: 30, lr: 5.91e-03, grad_scale: 16.0 +2023-02-06 13:27:35,466 INFO [optim.py:369] (0/4) 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 13:27:35,626 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100256.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:28:04,521 INFO [train.py:901] (0/4) Epoch 13, batch 3300, loss[loss=0.2297, simple_loss=0.3102, pruned_loss=0.0746, over 8282.00 frames. ], tot_loss[loss=0.2281, simple_loss=0.3056, pruned_loss=0.07529, over 1619796.37 frames. ], batch size: 23, lr: 5.90e-03, grad_scale: 16.0 +2023-02-06 13:28:04,763 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4513, 1.7215, 2.8772, 1.3021, 2.0147, 1.8208, 1.5078, 1.9010], + device='cuda:0'), covar=tensor([0.1782, 0.2199, 0.0631, 0.3932, 0.1592, 0.2840, 0.1936, 0.1937], + device='cuda:0'), in_proj_covar=tensor([0.0494, 0.0533, 0.0537, 0.0591, 0.0625, 0.0558, 0.0486, 0.0615], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 13:28:07,512 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100301.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:28:22,244 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100322.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:28:24,891 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100326.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:28:39,552 INFO [train.py:901] (0/4) Epoch 13, batch 3350, loss[loss=0.2005, simple_loss=0.2731, pruned_loss=0.06398, over 7799.00 frames. ], tot_loss[loss=0.2265, simple_loss=0.3043, pruned_loss=0.07431, over 1613646.49 frames. ], batch size: 19, lr: 5.90e-03, grad_scale: 16.0 +2023-02-06 13:28:39,753 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 13:28:54,366 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=100369.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 13:29:02,351 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3371, 2.5722, 1.5793, 2.0571, 1.9810, 1.2325, 1.8588, 2.0900], + device='cuda:0'), covar=tensor([0.1787, 0.0530, 0.1514, 0.0856, 0.0914, 0.1956, 0.1380, 0.1192], + device='cuda:0'), in_proj_covar=tensor([0.0343, 0.0236, 0.0319, 0.0300, 0.0298, 0.0322, 0.0340, 0.0303], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 13:29:13,503 INFO [train.py:901] (0/4) Epoch 13, batch 3400, loss[loss=0.2326, simple_loss=0.2891, pruned_loss=0.08805, over 7655.00 frames. ], tot_loss[loss=0.2276, simple_loss=0.3052, pruned_loss=0.07499, over 1617685.92 frames. ], batch size: 19, lr: 5.90e-03, grad_scale: 16.0 +2023-02-06 13:29:25,263 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100413.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:29:29,921 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100420.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:29:42,776 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100438.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:29:48,799 INFO [train.py:901] (0/4) Epoch 13, batch 3450, loss[loss=0.2052, simple_loss=0.2969, pruned_loss=0.05678, over 8451.00 frames. ], tot_loss[loss=0.2287, simple_loss=0.306, pruned_loss=0.07574, over 1615695.04 frames. ], batch size: 29, lr: 5.90e-03, grad_scale: 16.0 +2023-02-06 13:29:54,821 INFO [optim.py:369] (0/4) 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 13:30:07,207 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3334, 1.4754, 4.4336, 1.8914, 2.2694, 5.1366, 5.0737, 4.4064], + device='cuda:0'), covar=tensor([0.1081, 0.1759, 0.0273, 0.2049, 0.1295, 0.0175, 0.0292, 0.0531], + device='cuda:0'), in_proj_covar=tensor([0.0270, 0.0303, 0.0268, 0.0296, 0.0281, 0.0242, 0.0361, 0.0295], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 13:30:07,918 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0118, 1.3534, 1.5884, 1.1874, 1.0051, 1.3326, 1.8304, 1.4414], + device='cuda:0'), covar=tensor([0.0498, 0.1249, 0.1679, 0.1418, 0.0607, 0.1482, 0.0625, 0.0634], + device='cuda:0'), in_proj_covar=tensor([0.0098, 0.0151, 0.0190, 0.0157, 0.0101, 0.0162, 0.0113, 0.0138], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 13:30:14,288 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100484.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 13:30:23,411 INFO [train.py:901] (0/4) Epoch 13, batch 3500, loss[loss=0.2059, simple_loss=0.2964, pruned_loss=0.0577, over 8518.00 frames. ], tot_loss[loss=0.2283, simple_loss=0.3061, pruned_loss=0.07528, over 1617616.16 frames. ], batch size: 28, lr: 5.90e-03, grad_scale: 16.0 +2023-02-06 13:30:23,584 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8590, 1.6343, 3.1200, 1.3470, 2.1637, 3.3697, 3.4289, 2.8080], + device='cuda:0'), covar=tensor([0.1080, 0.1489, 0.0372, 0.2104, 0.0999, 0.0261, 0.0549, 0.0657], + device='cuda:0'), in_proj_covar=tensor([0.0270, 0.0303, 0.0269, 0.0296, 0.0281, 0.0242, 0.0362, 0.0296], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 13:30:45,069 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.1492, 1.2322, 4.2797, 1.6399, 3.7585, 3.5583, 3.8715, 3.7415], + device='cuda:0'), covar=tensor([0.0552, 0.4584, 0.0520, 0.3587, 0.1092, 0.0977, 0.0561, 0.0653], + device='cuda:0'), in_proj_covar=tensor([0.0516, 0.0593, 0.0613, 0.0557, 0.0632, 0.0545, 0.0533, 0.0593], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 13:30:53,023 WARNING [train.py:1067] (0/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-06 13:30:58,412 INFO [train.py:901] (0/4) Epoch 13, batch 3550, loss[loss=0.1637, simple_loss=0.2376, pruned_loss=0.04492, over 7422.00 frames. ], tot_loss[loss=0.2276, simple_loss=0.3047, pruned_loss=0.07527, over 1616079.87 frames. ], batch size: 17, lr: 5.90e-03, grad_scale: 16.0 +2023-02-06 13:31:04,442 INFO [optim.py:369] (0/4) 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 13:31:33,174 INFO [train.py:901] (0/4) Epoch 13, batch 3600, loss[loss=0.2514, simple_loss=0.3249, pruned_loss=0.08894, over 8592.00 frames. ], tot_loss[loss=0.2285, simple_loss=0.3051, pruned_loss=0.07592, over 1607984.13 frames. ], batch size: 31, lr: 5.89e-03, grad_scale: 16.0 +2023-02-06 13:31:35,338 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=100600.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:31:56,121 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0815, 1.7127, 1.4384, 1.6566, 1.3590, 1.2697, 1.3262, 1.4065], + device='cuda:0'), covar=tensor([0.1002, 0.0398, 0.0998, 0.0469, 0.0649, 0.1291, 0.0828, 0.0689], + device='cuda:0'), in_proj_covar=tensor([0.0341, 0.0233, 0.0317, 0.0297, 0.0296, 0.0322, 0.0339, 0.0301], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 13:32:08,079 INFO [train.py:901] (0/4) Epoch 13, batch 3650, loss[loss=0.1932, simple_loss=0.2678, pruned_loss=0.05931, over 7693.00 frames. ], tot_loss[loss=0.2276, simple_loss=0.3041, pruned_loss=0.07554, over 1605385.36 frames. ], batch size: 18, lr: 5.89e-03, grad_scale: 16.0 +2023-02-06 13:32:12,309 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9034, 1.6175, 1.6268, 1.5530, 1.0853, 1.5673, 1.7145, 1.7647], + device='cuda:0'), covar=tensor([0.0500, 0.0944, 0.1308, 0.1081, 0.0571, 0.1140, 0.0612, 0.0435], + device='cuda:0'), in_proj_covar=tensor([0.0099, 0.0152, 0.0191, 0.0158, 0.0102, 0.0163, 0.0115, 0.0139], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0008, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 13:32:14,110 INFO [optim.py:369] (0/4) 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 13:32:17,452 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 +2023-02-06 13:32:34,201 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.09 vs. limit=5.0 +2023-02-06 13:32:43,149 INFO [train.py:901] (0/4) Epoch 13, batch 3700, loss[loss=0.2027, simple_loss=0.2776, pruned_loss=0.0639, over 7796.00 frames. ], tot_loss[loss=0.2276, simple_loss=0.3038, pruned_loss=0.07566, over 1603934.29 frames. ], batch size: 19, lr: 5.89e-03, grad_scale: 8.0 +2023-02-06 13:32:52,912 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 +2023-02-06 13:32:55,259 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100715.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:32:57,154 WARNING [train.py:1067] (0/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-06 13:33:13,264 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100740.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 13:33:16,637 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5308, 2.6980, 1.9682, 2.1565, 2.0997, 1.5775, 1.9840, 2.1262], + device='cuda:0'), covar=tensor([0.1445, 0.0390, 0.0995, 0.0638, 0.0711, 0.1401, 0.1019, 0.0939], + device='cuda:0'), in_proj_covar=tensor([0.0344, 0.0234, 0.0318, 0.0299, 0.0300, 0.0322, 0.0340, 0.0303], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 13:33:17,851 INFO [train.py:901] (0/4) Epoch 13, batch 3750, loss[loss=0.2196, simple_loss=0.2963, pruned_loss=0.07149, over 8340.00 frames. ], tot_loss[loss=0.2278, simple_loss=0.3048, pruned_loss=0.07539, over 1613407.04 frames. ], batch size: 26, lr: 5.89e-03, grad_scale: 8.0 +2023-02-06 13:33:18,665 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100748.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 13:33:24,690 INFO [optim.py:369] (0/4) 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 13:33:30,125 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=100764.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:33:30,980 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100765.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 13:33:52,964 INFO [train.py:901] (0/4) Epoch 13, batch 3800, loss[loss=0.2671, simple_loss=0.328, pruned_loss=0.1031, over 7338.00 frames. ], tot_loss[loss=0.2285, simple_loss=0.3053, pruned_loss=0.07592, over 1614810.66 frames. ], batch size: 72, lr: 5.89e-03, grad_scale: 8.0 +2023-02-06 13:33:53,763 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.6123, 1.5155, 4.7013, 1.7446, 4.2470, 3.8944, 4.2539, 4.1432], + device='cuda:0'), covar=tensor([0.0463, 0.4255, 0.0512, 0.3429, 0.0901, 0.0918, 0.0504, 0.0531], + device='cuda:0'), in_proj_covar=tensor([0.0518, 0.0595, 0.0619, 0.0557, 0.0633, 0.0546, 0.0537, 0.0597], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 13:34:00,246 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 +2023-02-06 13:34:27,591 INFO [train.py:901] (0/4) Epoch 13, batch 3850, loss[loss=0.2477, simple_loss=0.3281, pruned_loss=0.0837, over 8522.00 frames. ], tot_loss[loss=0.2283, simple_loss=0.3052, pruned_loss=0.0757, over 1617614.67 frames. ], batch size: 28, lr: 5.89e-03, grad_scale: 8.0 +2023-02-06 13:34:34,506 INFO [optim.py:369] (0/4) 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 13:34:50,094 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100879.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:35:01,754 INFO [train.py:901] (0/4) Epoch 13, batch 3900, loss[loss=0.2499, simple_loss=0.319, pruned_loss=0.09041, over 7975.00 frames. ], tot_loss[loss=0.229, simple_loss=0.3058, pruned_loss=0.07612, over 1619477.51 frames. ], batch size: 21, lr: 5.89e-03, grad_scale: 8.0 +2023-02-06 13:35:01,758 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-06 13:35:33,298 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1887, 1.8437, 2.5346, 2.0563, 2.2952, 2.0739, 1.7380, 1.0207], + device='cuda:0'), covar=tensor([0.3836, 0.3576, 0.1250, 0.2490, 0.1890, 0.2271, 0.1661, 0.3994], + device='cuda:0'), in_proj_covar=tensor([0.0891, 0.0887, 0.0734, 0.0862, 0.0941, 0.0813, 0.0705, 0.0776], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 13:35:37,127 INFO [train.py:901] (0/4) Epoch 13, batch 3950, loss[loss=0.1997, simple_loss=0.2798, pruned_loss=0.05977, over 7807.00 frames. ], tot_loss[loss=0.229, simple_loss=0.3059, pruned_loss=0.07606, over 1615435.87 frames. ], batch size: 19, lr: 5.88e-03, grad_scale: 8.0 +2023-02-06 13:35:44,007 INFO [optim.py:369] (0/4) 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 13:35:54,141 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100971.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:36:10,578 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.2644, 1.1897, 3.3696, 1.0307, 2.9518, 2.8390, 3.0210, 2.9884], + device='cuda:0'), covar=tensor([0.0672, 0.3758, 0.0728, 0.3476, 0.1387, 0.0978, 0.0695, 0.0794], + device='cuda:0'), in_proj_covar=tensor([0.0513, 0.0588, 0.0612, 0.0551, 0.0630, 0.0543, 0.0533, 0.0593], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 13:36:11,866 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100996.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:36:12,360 INFO [train.py:901] (0/4) Epoch 13, batch 4000, loss[loss=0.206, simple_loss=0.289, pruned_loss=0.06151, over 8111.00 frames. ], tot_loss[loss=0.2293, simple_loss=0.3063, pruned_loss=0.07612, over 1613350.19 frames. ], batch size: 23, lr: 5.88e-03, grad_scale: 8.0 +2023-02-06 13:36:20,494 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.01 vs. limit=5.0 +2023-02-06 13:36:47,688 INFO [train.py:901] (0/4) Epoch 13, batch 4050, loss[loss=0.2244, simple_loss=0.3041, pruned_loss=0.07235, over 8473.00 frames. ], tot_loss[loss=0.2298, simple_loss=0.306, pruned_loss=0.07675, over 1607205.72 frames. ], batch size: 25, lr: 5.88e-03, grad_scale: 8.0 +2023-02-06 13:36:54,252 INFO [optim.py:369] (0/4) 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 13:37:18,350 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=101092.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 13:37:21,625 INFO [train.py:901] (0/4) Epoch 13, batch 4100, loss[loss=0.208, simple_loss=0.3006, pruned_loss=0.05765, over 8336.00 frames. ], tot_loss[loss=0.2286, simple_loss=0.3056, pruned_loss=0.0758, over 1612200.54 frames. ], batch size: 25, lr: 5.88e-03, grad_scale: 8.0 +2023-02-06 13:37:41,388 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=101125.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:37:47,947 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=101135.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:37:55,651 INFO [train.py:901] (0/4) Epoch 13, batch 4150, loss[loss=0.245, simple_loss=0.3286, pruned_loss=0.0807, over 7789.00 frames. ], tot_loss[loss=0.2291, simple_loss=0.306, pruned_loss=0.0761, over 1608652.47 frames. ], batch size: 19, lr: 5.88e-03, grad_scale: 8.0 +2023-02-06 13:38:02,996 INFO [optim.py:369] (0/4) 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 13:38:05,278 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101160.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:38:05,859 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.2540, 3.1398, 2.8847, 1.5880, 2.8889, 2.8989, 2.8646, 2.6899], + device='cuda:0'), covar=tensor([0.1183, 0.0858, 0.1319, 0.4391, 0.1048, 0.1329, 0.1704, 0.1290], + device='cuda:0'), in_proj_covar=tensor([0.0479, 0.0391, 0.0401, 0.0500, 0.0394, 0.0395, 0.0389, 0.0344], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 13:38:06,624 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9366, 1.2361, 1.5331, 1.1337, 0.9523, 1.2814, 1.6495, 1.5641], + device='cuda:0'), covar=tensor([0.0542, 0.1748, 0.2374, 0.1918, 0.0682, 0.2054, 0.0774, 0.0719], + device='cuda:0'), in_proj_covar=tensor([0.0099, 0.0153, 0.0192, 0.0157, 0.0102, 0.0164, 0.0116, 0.0139], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 13:38:30,234 INFO [train.py:901] (0/4) Epoch 13, batch 4200, loss[loss=0.2506, simple_loss=0.3181, pruned_loss=0.09157, over 6832.00 frames. ], tot_loss[loss=0.2285, simple_loss=0.3054, pruned_loss=0.07582, over 1610248.05 frames. ], batch size: 71, lr: 5.88e-03, grad_scale: 8.0 +2023-02-06 13:38:37,978 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=101207.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 13:38:39,773 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 +2023-02-06 13:38:54,609 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-06 13:39:05,863 INFO [train.py:901] (0/4) Epoch 13, batch 4250, loss[loss=0.2164, simple_loss=0.2803, pruned_loss=0.07625, over 7515.00 frames. ], tot_loss[loss=0.2277, simple_loss=0.3046, pruned_loss=0.07545, over 1608795.09 frames. ], batch size: 18, lr: 5.88e-03, grad_scale: 8.0 +2023-02-06 13:39:12,480 INFO [optim.py:369] (0/4) 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 13:39:16,506 WARNING [train.py:1067] (0/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-06 13:39:40,003 INFO [train.py:901] (0/4) Epoch 13, batch 4300, loss[loss=0.1975, simple_loss=0.2683, pruned_loss=0.06339, over 7541.00 frames. ], tot_loss[loss=0.2255, simple_loss=0.3027, pruned_loss=0.07415, over 1607466.64 frames. ], batch size: 18, lr: 5.87e-03, grad_scale: 8.0 +2023-02-06 13:39:58,268 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-06 13:40:01,018 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 +2023-02-06 13:40:14,572 INFO [train.py:901] (0/4) Epoch 13, batch 4350, loss[loss=0.289, simple_loss=0.3478, pruned_loss=0.1151, over 6816.00 frames. ], tot_loss[loss=0.2269, simple_loss=0.3042, pruned_loss=0.07485, over 1612344.32 frames. ], batch size: 71, lr: 5.87e-03, grad_scale: 8.0 +2023-02-06 13:40:14,795 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5758, 1.6354, 2.0935, 1.5458, 1.1781, 2.1101, 0.2680, 1.4002], + device='cuda:0'), covar=tensor([0.2339, 0.1658, 0.0419, 0.1359, 0.3825, 0.0482, 0.3068, 0.1424], + device='cuda:0'), in_proj_covar=tensor([0.0171, 0.0176, 0.0107, 0.0219, 0.0256, 0.0111, 0.0164, 0.0170], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 13:40:21,335 INFO [optim.py:369] (0/4) 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 13:40:40,928 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7084, 1.5243, 3.3628, 1.3873, 2.3412, 3.6166, 3.6749, 3.0730], + device='cuda:0'), covar=tensor([0.1140, 0.1539, 0.0310, 0.2157, 0.0852, 0.0235, 0.0446, 0.0575], + device='cuda:0'), in_proj_covar=tensor([0.0270, 0.0302, 0.0266, 0.0296, 0.0278, 0.0239, 0.0360, 0.0296], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 13:40:49,418 INFO [train.py:901] (0/4) Epoch 13, batch 4400, loss[loss=0.1703, simple_loss=0.2528, pruned_loss=0.0439, over 7698.00 frames. ], tot_loss[loss=0.2265, simple_loss=0.3035, pruned_loss=0.07477, over 1604429.74 frames. ], batch size: 18, lr: 5.87e-03, grad_scale: 8.0 +2023-02-06 13:40:49,429 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-06 13:41:23,440 INFO [train.py:901] (0/4) Epoch 13, batch 4450, loss[loss=0.217, simple_loss=0.2968, pruned_loss=0.06858, over 8704.00 frames. ], tot_loss[loss=0.2273, simple_loss=0.3044, pruned_loss=0.07512, over 1609527.42 frames. ], batch size: 39, lr: 5.87e-03, grad_scale: 8.0 +2023-02-06 13:41:28,826 WARNING [train.py:1067] (0/4) 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] (0/4) 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] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=101463.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 13:41:38,577 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=101469.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:41:52,065 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101488.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 13:41:57,473 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([0.9565, 1.6474, 1.4272, 1.5887, 1.3438, 1.2320, 1.1914, 1.3327], + device='cuda:0'), covar=tensor([0.1057, 0.0405, 0.1059, 0.0497, 0.0675, 0.1285, 0.0886, 0.0658], + device='cuda:0'), in_proj_covar=tensor([0.0343, 0.0233, 0.0314, 0.0295, 0.0298, 0.0319, 0.0338, 0.0302], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 13:41:57,884 INFO [train.py:901] (0/4) Epoch 13, batch 4500, loss[loss=0.2503, simple_loss=0.3269, pruned_loss=0.08682, over 8362.00 frames. ], tot_loss[loss=0.2284, simple_loss=0.3056, pruned_loss=0.07554, over 1613750.50 frames. ], batch size: 24, lr: 5.87e-03, grad_scale: 8.0 +2023-02-06 13:42:11,518 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5500, 2.1178, 3.4160, 1.3168, 2.4555, 1.9730, 1.6212, 2.5163], + device='cuda:0'), covar=tensor([0.1707, 0.2110, 0.0735, 0.3991, 0.1646, 0.2833, 0.1914, 0.1984], + device='cuda:0'), in_proj_covar=tensor([0.0494, 0.0535, 0.0537, 0.0590, 0.0626, 0.0562, 0.0485, 0.0617], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 13:42:22,189 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-06 13:42:33,108 INFO [train.py:901] (0/4) Epoch 13, batch 4550, loss[loss=0.2164, simple_loss=0.3016, pruned_loss=0.06561, over 8466.00 frames. ], tot_loss[loss=0.2276, simple_loss=0.305, pruned_loss=0.07514, over 1616834.48 frames. ], batch size: 27, lr: 5.87e-03, grad_scale: 8.0 +2023-02-06 13:42:39,891 INFO [optim.py:369] (0/4) 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 13:42:58,926 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=101584.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:43:08,200 INFO [train.py:901] (0/4) Epoch 13, batch 4600, loss[loss=0.2085, simple_loss=0.2919, pruned_loss=0.06254, over 7657.00 frames. ], tot_loss[loss=0.2272, simple_loss=0.3047, pruned_loss=0.07485, over 1614682.04 frames. ], batch size: 19, lr: 5.87e-03, grad_scale: 8.0 +2023-02-06 13:43:42,608 INFO [train.py:901] (0/4) Epoch 13, batch 4650, loss[loss=0.1913, simple_loss=0.279, pruned_loss=0.05179, over 8527.00 frames. ], tot_loss[loss=0.2274, simple_loss=0.3047, pruned_loss=0.07502, over 1614344.06 frames. ], batch size: 34, lr: 5.86e-03, grad_scale: 8.0 +2023-02-06 13:43:49,456 INFO [optim.py:369] (0/4) 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 13:44:01,145 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0760, 1.6385, 1.3655, 1.6259, 1.4269, 1.1759, 1.3290, 1.3302], + device='cuda:0'), covar=tensor([0.0979, 0.0434, 0.1149, 0.0473, 0.0659, 0.1361, 0.0769, 0.0622], + device='cuda:0'), in_proj_covar=tensor([0.0345, 0.0235, 0.0316, 0.0295, 0.0300, 0.0321, 0.0340, 0.0304], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 13:44:17,683 INFO [train.py:901] (0/4) Epoch 13, batch 4700, loss[loss=0.1968, simple_loss=0.268, pruned_loss=0.06277, over 7802.00 frames. ], tot_loss[loss=0.2274, simple_loss=0.3043, pruned_loss=0.07523, over 1614354.33 frames. ], batch size: 19, lr: 5.86e-03, grad_scale: 8.0 +2023-02-06 13:44:21,749 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=101702.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:44:52,091 INFO [train.py:901] (0/4) Epoch 13, batch 4750, loss[loss=0.2417, simple_loss=0.2904, pruned_loss=0.09647, over 7447.00 frames. ], tot_loss[loss=0.2263, simple_loss=0.3034, pruned_loss=0.07459, over 1613734.69 frames. ], batch size: 17, lr: 5.86e-03, grad_scale: 8.0 +2023-02-06 13:44:59,502 INFO [optim.py:369] (0/4) 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 13:45:21,976 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-06 13:45:24,391 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-06 13:45:27,102 INFO [train.py:901] (0/4) Epoch 13, batch 4800, loss[loss=0.1825, simple_loss=0.2639, pruned_loss=0.05058, over 7709.00 frames. ], tot_loss[loss=0.2253, simple_loss=0.3028, pruned_loss=0.07392, over 1611683.22 frames. ], batch size: 18, lr: 5.86e-03, grad_scale: 8.0 +2023-02-06 13:45:57,600 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 13, batch 4850, loss[loss=0.2135, simple_loss=0.2886, pruned_loss=0.06916, over 8244.00 frames. ], tot_loss[loss=0.2256, simple_loss=0.3031, pruned_loss=0.07405, over 1613760.67 frames. ], batch size: 22, lr: 5.86e-03, grad_scale: 8.0 +2023-02-06 13:46:08,646 INFO [optim.py:369] (0/4) 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 13:46:14,079 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-06 13:46:14,953 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 13, batch 4900, loss[loss=0.2941, simple_loss=0.351, pruned_loss=0.1186, over 7303.00 frames. ], tot_loss[loss=0.2266, simple_loss=0.3038, pruned_loss=0.07464, over 1611716.67 frames. ], batch size: 72, lr: 5.86e-03, grad_scale: 8.0 +2023-02-06 13:46:52,169 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 +2023-02-06 13:47:06,369 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=101938.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 13:47:11,284 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-06 13:47:12,291 INFO [train.py:901] (0/4) Epoch 13, batch 4950, loss[loss=0.2501, simple_loss=0.3242, pruned_loss=0.08803, over 8419.00 frames. ], tot_loss[loss=0.2257, simple_loss=0.303, pruned_loss=0.0742, over 1614414.39 frames. ], batch size: 29, lr: 5.86e-03, grad_scale: 8.0 +2023-02-06 13:47:15,207 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5675, 2.8016, 2.0444, 2.3929, 2.3217, 1.5944, 2.0885, 2.1739], + device='cuda:0'), covar=tensor([0.1384, 0.0333, 0.0879, 0.0587, 0.0532, 0.1299, 0.0966, 0.0959], + device='cuda:0'), in_proj_covar=tensor([0.0343, 0.0233, 0.0314, 0.0293, 0.0296, 0.0321, 0.0339, 0.0302], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 13:47:19,104 INFO [optim.py:369] (0/4) 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 13:47:30,002 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.9977, 1.8359, 1.9015, 1.7765, 1.0817, 1.8395, 2.2404, 2.2225], + device='cuda:0'), covar=tensor([0.0421, 0.1111, 0.1586, 0.1323, 0.0605, 0.1362, 0.0608, 0.0525], + device='cuda:0'), in_proj_covar=tensor([0.0098, 0.0152, 0.0191, 0.0157, 0.0102, 0.0163, 0.0114, 0.0138], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 13:47:46,757 INFO [train.py:901] (0/4) Epoch 13, batch 5000, loss[loss=0.2075, simple_loss=0.298, pruned_loss=0.05849, over 8450.00 frames. ], tot_loss[loss=0.2251, simple_loss=0.3021, pruned_loss=0.07399, over 1610292.15 frames. ], batch size: 27, lr: 5.85e-03, grad_scale: 8.0 +2023-02-06 13:47:48,887 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-102000.pt +2023-02-06 13:47:53,363 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=102005.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:48:22,480 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 13, batch 5050, loss[loss=0.264, simple_loss=0.337, pruned_loss=0.09545, over 8106.00 frames. ], tot_loss[loss=0.2249, simple_loss=0.302, pruned_loss=0.07394, over 1613280.14 frames. ], batch size: 23, lr: 5.85e-03, grad_scale: 8.0 +2023-02-06 13:48:29,968 INFO [optim.py:369] (0/4) 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 13:48:54,016 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-06 13:48:56,649 INFO [train.py:901] (0/4) Epoch 13, batch 5100, loss[loss=0.2287, simple_loss=0.3173, pruned_loss=0.07008, over 8439.00 frames. ], tot_loss[loss=0.225, simple_loss=0.3023, pruned_loss=0.07382, over 1612710.13 frames. ], batch size: 27, lr: 5.85e-03, grad_scale: 8.0 +2023-02-06 13:49:31,560 INFO [train.py:901] (0/4) Epoch 13, batch 5150, loss[loss=0.196, simple_loss=0.281, pruned_loss=0.05546, over 8090.00 frames. ], tot_loss[loss=0.2249, simple_loss=0.3024, pruned_loss=0.07367, over 1617301.89 frames. ], batch size: 21, lr: 5.85e-03, grad_scale: 8.0 +2023-02-06 13:49:38,302 INFO [optim.py:369] (0/4) 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 13:49:41,790 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=102161.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:50:06,640 INFO [train.py:901] (0/4) Epoch 13, batch 5200, loss[loss=0.2155, simple_loss=0.2908, pruned_loss=0.07008, over 8075.00 frames. ], tot_loss[loss=0.2251, simple_loss=0.3024, pruned_loss=0.07393, over 1616718.60 frames. ], batch size: 21, lr: 5.85e-03, grad_scale: 8.0 +2023-02-06 13:50:41,852 INFO [train.py:901] (0/4) Epoch 13, batch 5250, loss[loss=0.224, simple_loss=0.3063, pruned_loss=0.07079, over 8448.00 frames. ], tot_loss[loss=0.227, simple_loss=0.3042, pruned_loss=0.0749, over 1619113.25 frames. ], batch size: 27, lr: 5.85e-03, grad_scale: 8.0 +2023-02-06 13:50:48,569 INFO [optim.py:369] (0/4) 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] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.92 vs. limit=5.0 +2023-02-06 13:50:53,900 WARNING [train.py:1067] (0/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-06 13:51:06,836 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=102282.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 13:51:16,473 INFO [train.py:901] (0/4) Epoch 13, batch 5300, loss[loss=0.2125, simple_loss=0.2976, pruned_loss=0.06375, over 7813.00 frames. ], tot_loss[loss=0.225, simple_loss=0.3023, pruned_loss=0.07389, over 1614216.83 frames. ], batch size: 20, lr: 5.85e-03, grad_scale: 8.0 +2023-02-06 13:51:51,011 INFO [train.py:901] (0/4) Epoch 13, batch 5350, loss[loss=0.2249, simple_loss=0.2917, pruned_loss=0.07903, over 7656.00 frames. ], tot_loss[loss=0.2246, simple_loss=0.3019, pruned_loss=0.07369, over 1607574.26 frames. ], batch size: 19, lr: 5.84e-03, grad_scale: 8.0 +2023-02-06 13:51:52,455 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=102349.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:51:52,583 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6424, 2.8267, 1.8316, 2.1979, 2.2918, 1.5497, 2.1151, 2.2111], + device='cuda:0'), covar=tensor([0.1271, 0.0324, 0.1089, 0.0667, 0.0728, 0.1417, 0.1015, 0.0898], + device='cuda:0'), in_proj_covar=tensor([0.0347, 0.0236, 0.0318, 0.0297, 0.0298, 0.0324, 0.0345, 0.0304], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 13:51:57,801 INFO [optim.py:369] (0/4) 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 13:52:08,203 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.55 vs. limit=2.0 +2023-02-06 13:52:26,059 INFO [train.py:901] (0/4) Epoch 13, batch 5400, loss[loss=0.2401, simple_loss=0.3064, pruned_loss=0.08689, over 7923.00 frames. ], tot_loss[loss=0.2263, simple_loss=0.3034, pruned_loss=0.07457, over 1609920.53 frames. ], batch size: 20, lr: 5.84e-03, grad_scale: 8.0 +2023-02-06 13:52:26,259 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=102397.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 13:52:40,296 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=102417.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:52:56,871 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=102442.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:53:00,134 INFO [train.py:901] (0/4) Epoch 13, batch 5450, loss[loss=0.1828, simple_loss=0.2526, pruned_loss=0.05653, over 7428.00 frames. ], tot_loss[loss=0.2271, simple_loss=0.3039, pruned_loss=0.07518, over 1607492.10 frames. ], batch size: 17, lr: 5.84e-03, grad_scale: 8.0 +2023-02-06 13:53:07,656 INFO [optim.py:369] (0/4) 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] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=102464.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:53:34,968 INFO [train.py:901] (0/4) Epoch 13, batch 5500, loss[loss=0.2657, simple_loss=0.3445, pruned_loss=0.09343, over 8469.00 frames. ], tot_loss[loss=0.2273, simple_loss=0.3041, pruned_loss=0.07528, over 1606125.39 frames. ], batch size: 27, lr: 5.84e-03, grad_scale: 8.0 +2023-02-06 13:53:41,598 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-06 13:54:03,089 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0284, 1.5443, 3.2654, 1.3607, 2.2091, 3.6308, 3.7064, 3.0859], + device='cuda:0'), covar=tensor([0.0991, 0.1530, 0.0325, 0.2171, 0.1027, 0.0221, 0.0409, 0.0577], + device='cuda:0'), in_proj_covar=tensor([0.0268, 0.0300, 0.0265, 0.0296, 0.0277, 0.0237, 0.0360, 0.0295], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 13:54:09,471 INFO [train.py:901] (0/4) Epoch 13, batch 5550, loss[loss=0.1991, simple_loss=0.2863, pruned_loss=0.056, over 8469.00 frames. ], tot_loss[loss=0.2262, simple_loss=0.3029, pruned_loss=0.07479, over 1603379.41 frames. ], batch size: 25, lr: 5.84e-03, grad_scale: 8.0 +2023-02-06 13:54:15,944 INFO [optim.py:369] (0/4) 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 13:54:43,200 INFO [train.py:901] (0/4) Epoch 13, batch 5600, loss[loss=0.2299, simple_loss=0.3219, pruned_loss=0.06898, over 8461.00 frames. ], tot_loss[loss=0.2271, simple_loss=0.3035, pruned_loss=0.07534, over 1602312.74 frames. ], batch size: 29, lr: 5.84e-03, grad_scale: 8.0 +2023-02-06 13:54:44,041 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4997, 1.5084, 4.2006, 1.8104, 2.2068, 4.8826, 4.9223, 4.1368], + device='cuda:0'), covar=tensor([0.0927, 0.1782, 0.0303, 0.2064, 0.1223, 0.0167, 0.0332, 0.0565], + device='cuda:0'), in_proj_covar=tensor([0.0269, 0.0302, 0.0266, 0.0296, 0.0278, 0.0239, 0.0360, 0.0296], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 13:55:04,659 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7130, 1.9636, 2.1190, 1.1846, 2.1802, 1.5371, 0.5338, 1.9392], + device='cuda:0'), covar=tensor([0.0339, 0.0221, 0.0160, 0.0392, 0.0269, 0.0520, 0.0572, 0.0165], + device='cuda:0'), in_proj_covar=tensor([0.0406, 0.0341, 0.0293, 0.0401, 0.0332, 0.0490, 0.0368, 0.0373], + device='cuda:0'), 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:0') +2023-02-06 13:55:18,283 INFO [train.py:901] (0/4) Epoch 13, batch 5650, loss[loss=0.1995, simple_loss=0.2767, pruned_loss=0.06116, over 7927.00 frames. ], tot_loss[loss=0.2271, simple_loss=0.3037, pruned_loss=0.07519, over 1604071.12 frames. ], batch size: 20, lr: 5.84e-03, grad_scale: 8.0 +2023-02-06 13:55:22,471 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=102653.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 13:55:24,852 INFO [optim.py:369] (0/4) 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 13:55:39,200 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=102678.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 13:55:43,637 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 13, batch 5700, loss[loss=0.1751, simple_loss=0.2519, pruned_loss=0.04909, over 7426.00 frames. ], tot_loss[loss=0.2302, simple_loss=0.306, pruned_loss=0.07716, over 1600076.57 frames. ], batch size: 17, lr: 5.83e-03, grad_scale: 16.0 +2023-02-06 13:56:08,739 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=102720.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:56:25,984 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=102745.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:56:27,757 INFO [train.py:901] (0/4) Epoch 13, batch 5750, loss[loss=0.2088, simple_loss=0.2811, pruned_loss=0.06821, over 7647.00 frames. ], tot_loss[loss=0.2302, simple_loss=0.3066, pruned_loss=0.07692, over 1606046.40 frames. ], batch size: 19, lr: 5.83e-03, grad_scale: 16.0 +2023-02-06 13:56:34,419 INFO [optim.py:369] (0/4) 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 13:56:35,995 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4084, 1.8409, 2.9678, 1.1563, 2.3298, 1.8516, 1.5448, 2.1447], + device='cuda:0'), covar=tensor([0.1810, 0.2201, 0.0796, 0.4270, 0.1584, 0.3036, 0.2046, 0.2197], + device='cuda:0'), in_proj_covar=tensor([0.0491, 0.0529, 0.0530, 0.0585, 0.0617, 0.0557, 0.0481, 0.0608], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 13:56:37,296 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3248, 2.6718, 1.7884, 2.1157, 2.0639, 1.4130, 1.8493, 2.0413], + device='cuda:0'), covar=tensor([0.1428, 0.0318, 0.1065, 0.0613, 0.0656, 0.1435, 0.1009, 0.0845], + device='cuda:0'), in_proj_covar=tensor([0.0351, 0.0237, 0.0323, 0.0301, 0.0300, 0.0327, 0.0346, 0.0308], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 13:56:47,306 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-06 13:57:01,392 INFO [train.py:901] (0/4) Epoch 13, batch 5800, loss[loss=0.2246, simple_loss=0.2919, pruned_loss=0.07864, over 8494.00 frames. ], tot_loss[loss=0.2288, simple_loss=0.3055, pruned_loss=0.07606, over 1608148.13 frames. ], batch size: 26, lr: 5.83e-03, grad_scale: 16.0 +2023-02-06 13:57:28,961 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6690, 1.6228, 2.1313, 1.5275, 1.1479, 2.1156, 0.2936, 1.3350], + device='cuda:0'), covar=tensor([0.2452, 0.1524, 0.0417, 0.1508, 0.3693, 0.0358, 0.2747, 0.1662], + device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0174, 0.0106, 0.0220, 0.0255, 0.0110, 0.0164, 0.0172], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 13:57:36,623 INFO [train.py:901] (0/4) Epoch 13, batch 5850, loss[loss=0.2128, simple_loss=0.2864, pruned_loss=0.06961, over 7934.00 frames. ], tot_loss[loss=0.2282, simple_loss=0.3047, pruned_loss=0.0759, over 1605469.29 frames. ], batch size: 20, lr: 5.83e-03, grad_scale: 16.0 +2023-02-06 13:57:36,815 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8227, 1.8343, 1.7999, 2.6071, 1.1653, 1.5124, 1.8627, 2.1394], + device='cuda:0'), covar=tensor([0.0787, 0.1044, 0.1024, 0.0402, 0.1220, 0.1469, 0.0855, 0.0770], + device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0210, 0.0254, 0.0212, 0.0216, 0.0254, 0.0256, 0.0218], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 13:57:43,164 INFO [optim.py:369] (0/4) 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 13:57:52,582 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=102869.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:58:11,316 INFO [train.py:901] (0/4) Epoch 13, batch 5900, loss[loss=0.239, simple_loss=0.3292, pruned_loss=0.07437, over 8359.00 frames. ], tot_loss[loss=0.2281, simple_loss=0.3048, pruned_loss=0.07572, over 1611368.28 frames. ], batch size: 49, lr: 5.83e-03, grad_scale: 16.0 +2023-02-06 13:58:16,829 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=102905.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 13:58:29,671 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0754, 1.3707, 1.5675, 1.3568, 0.9283, 1.4278, 1.6533, 1.5109], + device='cuda:0'), covar=tensor([0.0484, 0.1307, 0.1701, 0.1379, 0.0628, 0.1533, 0.0686, 0.0635], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0152, 0.0191, 0.0157, 0.0101, 0.0162, 0.0113, 0.0138], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0008, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 13:58:42,653 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.85 vs. limit=5.0 +2023-02-06 13:58:46,262 INFO [train.py:901] (0/4) Epoch 13, batch 5950, loss[loss=0.2907, simple_loss=0.3476, pruned_loss=0.1169, over 6875.00 frames. ], tot_loss[loss=0.2283, simple_loss=0.3054, pruned_loss=0.07556, over 1614592.90 frames. ], batch size: 71, lr: 5.83e-03, grad_scale: 16.0 +2023-02-06 13:58:52,886 INFO [optim.py:369] (0/4) 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 13:58:58,993 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0786, 2.6432, 3.0600, 1.3273, 3.1461, 1.7606, 1.5768, 2.0829], + device='cuda:0'), covar=tensor([0.0646, 0.0282, 0.0184, 0.0629, 0.0421, 0.0744, 0.0691, 0.0426], + device='cuda:0'), in_proj_covar=tensor([0.0410, 0.0346, 0.0296, 0.0406, 0.0334, 0.0496, 0.0372, 0.0376], + device='cuda:0'), out_proj_covar=tensor([1.1459e-04, 9.4281e-05, 8.0591e-05, 1.1148e-04, 9.2291e-05, 1.4726e-04, + 1.0437e-04, 1.0401e-04], device='cuda:0') +2023-02-06 13:59:21,495 INFO [train.py:901] (0/4) Epoch 13, batch 6000, loss[loss=0.1973, simple_loss=0.2747, pruned_loss=0.05998, over 7938.00 frames. ], tot_loss[loss=0.2276, simple_loss=0.3053, pruned_loss=0.07501, over 1614084.11 frames. ], batch size: 20, lr: 5.83e-03, grad_scale: 16.0 +2023-02-06 13:59:21,496 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 13:59:36,610 INFO [train.py:935] (0/4) Epoch 13, validation: loss=0.1836, simple_loss=0.2836, pruned_loss=0.04176, over 944034.00 frames. +2023-02-06 13:59:36,611 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-06 13:59:41,487 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 +2023-02-06 14:00:03,143 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.5587, 1.3097, 4.6700, 1.7469, 4.0686, 3.8817, 4.2421, 4.1113], + device='cuda:0'), covar=tensor([0.0479, 0.4792, 0.0515, 0.3909, 0.1150, 0.0990, 0.0526, 0.0619], + device='cuda:0'), in_proj_covar=tensor([0.0512, 0.0579, 0.0603, 0.0552, 0.0623, 0.0529, 0.0523, 0.0582], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 14:00:10,254 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 +2023-02-06 14:00:11,023 INFO [train.py:901] (0/4) Epoch 13, batch 6050, loss[loss=0.1896, simple_loss=0.264, pruned_loss=0.05757, over 7704.00 frames. ], tot_loss[loss=0.2262, simple_loss=0.3038, pruned_loss=0.07435, over 1613629.93 frames. ], batch size: 18, lr: 5.82e-03, grad_scale: 16.0 +2023-02-06 14:00:18,297 INFO [optim.py:369] (0/4) 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 14:00:45,649 INFO [train.py:901] (0/4) Epoch 13, batch 6100, loss[loss=0.196, simple_loss=0.2802, pruned_loss=0.05595, over 8358.00 frames. ], tot_loss[loss=0.2269, simple_loss=0.3045, pruned_loss=0.07466, over 1610807.96 frames. ], batch size: 24, lr: 5.82e-03, grad_scale: 16.0 +2023-02-06 14:00:52,616 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6409, 1.8951, 1.9613, 1.2381, 2.0765, 1.4537, 0.4630, 1.8435], + device='cuda:0'), covar=tensor([0.0380, 0.0246, 0.0196, 0.0347, 0.0241, 0.0678, 0.0583, 0.0180], + device='cuda:0'), in_proj_covar=tensor([0.0409, 0.0343, 0.0296, 0.0405, 0.0333, 0.0496, 0.0372, 0.0376], + device='cuda:0'), out_proj_covar=tensor([1.1437e-04, 9.3582e-05, 8.0512e-05, 1.1142e-04, 9.2098e-05, 1.4711e-04, + 1.0433e-04, 1.0395e-04], device='cuda:0') +2023-02-06 14:00:57,237 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0574, 1.7007, 3.4483, 1.6283, 2.3487, 3.8316, 3.8366, 3.2472], + device='cuda:0'), covar=tensor([0.1045, 0.1516, 0.0306, 0.1843, 0.0980, 0.0205, 0.0468, 0.0564], + device='cuda:0'), in_proj_covar=tensor([0.0265, 0.0299, 0.0262, 0.0293, 0.0274, 0.0237, 0.0357, 0.0292], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 14:00:58,574 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103116.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:01:14,154 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-06 14:01:19,569 INFO [train.py:901] (0/4) Epoch 13, batch 6150, loss[loss=0.2183, simple_loss=0.2807, pruned_loss=0.07796, over 7432.00 frames. ], tot_loss[loss=0.2276, simple_loss=0.3048, pruned_loss=0.07521, over 1610591.95 frames. ], batch size: 17, lr: 5.82e-03, grad_scale: 16.0 +2023-02-06 14:01:21,548 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103150.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:01:26,197 INFO [optim.py:369] (0/4) 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 14:01:55,077 INFO [train.py:901] (0/4) Epoch 13, batch 6200, loss[loss=0.2387, simple_loss=0.3221, pruned_loss=0.07767, over 8250.00 frames. ], tot_loss[loss=0.2266, simple_loss=0.3037, pruned_loss=0.07473, over 1611169.06 frames. ], batch size: 24, lr: 5.82e-03, grad_scale: 16.0 +2023-02-06 14:02:06,142 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=103213.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:02:30,996 INFO [train.py:901] (0/4) Epoch 13, batch 6250, loss[loss=0.2084, simple_loss=0.2933, pruned_loss=0.06172, over 8200.00 frames. ], tot_loss[loss=0.2246, simple_loss=0.3023, pruned_loss=0.0735, over 1611327.29 frames. ], batch size: 23, lr: 5.82e-03, grad_scale: 16.0 +2023-02-06 14:02:32,350 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=103249.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:02:32,480 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0494, 4.0029, 2.4500, 2.9430, 3.0120, 2.2608, 2.9697, 3.1371], + device='cuda:0'), covar=tensor([0.1577, 0.0317, 0.1039, 0.0687, 0.0705, 0.1280, 0.0940, 0.0938], + device='cuda:0'), in_proj_covar=tensor([0.0349, 0.0239, 0.0325, 0.0299, 0.0300, 0.0326, 0.0344, 0.0309], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 14:02:37,848 INFO [optim.py:369] (0/4) 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 14:02:57,950 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5439, 2.3875, 1.6037, 2.1194, 2.0344, 1.3565, 1.9441, 2.1088], + device='cuda:0'), covar=tensor([0.1199, 0.0364, 0.1145, 0.0554, 0.0677, 0.1419, 0.0803, 0.0677], + device='cuda:0'), in_proj_covar=tensor([0.0348, 0.0238, 0.0323, 0.0298, 0.0299, 0.0324, 0.0341, 0.0307], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 14:03:06,053 INFO [train.py:901] (0/4) Epoch 13, batch 6300, loss[loss=0.2589, simple_loss=0.3377, pruned_loss=0.09004, over 8619.00 frames. ], tot_loss[loss=0.2241, simple_loss=0.3015, pruned_loss=0.07338, over 1610376.35 frames. ], batch size: 39, lr: 5.82e-03, grad_scale: 16.0 +2023-02-06 14:03:27,927 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103328.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:03:40,550 INFO [train.py:901] (0/4) Epoch 13, batch 6350, loss[loss=0.2184, simple_loss=0.2891, pruned_loss=0.07385, over 8066.00 frames. ], tot_loss[loss=0.224, simple_loss=0.3011, pruned_loss=0.07346, over 1608587.22 frames. ], batch size: 21, lr: 5.82e-03, grad_scale: 16.0 +2023-02-06 14:03:48,226 INFO [optim.py:369] (0/4) 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 14:03:52,994 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103364.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:04:06,344 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103384.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 14:04:14,868 INFO [train.py:901] (0/4) Epoch 13, batch 6400, loss[loss=0.1825, simple_loss=0.2656, pruned_loss=0.0497, over 8087.00 frames. ], tot_loss[loss=0.2244, simple_loss=0.3011, pruned_loss=0.07385, over 1601719.39 frames. ], batch size: 21, lr: 5.81e-03, grad_scale: 16.0 +2023-02-06 14:04:49,441 INFO [train.py:901] (0/4) Epoch 13, batch 6450, loss[loss=0.2228, simple_loss=0.2887, pruned_loss=0.07843, over 7654.00 frames. ], tot_loss[loss=0.2249, simple_loss=0.3015, pruned_loss=0.07415, over 1600087.89 frames. ], batch size: 19, lr: 5.81e-03, grad_scale: 16.0 +2023-02-06 14:04:56,176 INFO [optim.py:369] (0/4) 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 14:04:58,224 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=103460.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:05:13,646 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2412, 1.2322, 1.4793, 1.2041, 0.7215, 1.2771, 1.1979, 1.1987], + device='cuda:0'), covar=tensor([0.0563, 0.1280, 0.1696, 0.1413, 0.0575, 0.1534, 0.0681, 0.0651], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0150, 0.0189, 0.0155, 0.0100, 0.0161, 0.0113, 0.0137], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0008, 0.0006, 0.0008, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 14:05:22,245 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=103494.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:05:24,239 INFO [train.py:901] (0/4) Epoch 13, batch 6500, loss[loss=0.2246, simple_loss=0.3077, pruned_loss=0.07075, over 8498.00 frames. ], tot_loss[loss=0.2282, simple_loss=0.3045, pruned_loss=0.07595, over 1604480.84 frames. ], batch size: 28, lr: 5.81e-03, grad_scale: 16.0 +2023-02-06 14:05:58,666 INFO [train.py:901] (0/4) Epoch 13, batch 6550, loss[loss=0.2086, simple_loss=0.2754, pruned_loss=0.07093, over 7696.00 frames. ], tot_loss[loss=0.2298, simple_loss=0.3061, pruned_loss=0.07676, over 1612190.04 frames. ], batch size: 18, lr: 5.81e-03, grad_scale: 16.0 +2023-02-06 14:06:05,490 INFO [optim.py:369] (0/4) 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 14:06:18,321 INFO [zipformer.py:1185] (0/4) 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] (0/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-06 14:06:24,388 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 13, batch 6600, loss[loss=0.242, simple_loss=0.3168, pruned_loss=0.0836, over 8667.00 frames. ], tot_loss[loss=0.2294, simple_loss=0.3058, pruned_loss=0.07649, over 1612229.98 frames. ], batch size: 39, lr: 5.81e-03, grad_scale: 16.0 +2023-02-06 14:06:42,475 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103609.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:06:42,500 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103609.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:06:43,681 WARNING [train.py:1067] (0/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-06 14:06:49,849 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103620.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:06:52,145 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 +2023-02-06 14:06:54,802 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.6569, 2.2286, 3.5954, 2.5703, 3.1030, 2.4570, 2.1017, 1.7988], + device='cuda:0'), covar=tensor([0.4231, 0.4762, 0.1288, 0.2966, 0.2097, 0.2257, 0.1584, 0.4711], + device='cuda:0'), in_proj_covar=tensor([0.0897, 0.0889, 0.0743, 0.0868, 0.0948, 0.0824, 0.0708, 0.0776], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 14:07:07,861 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103645.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:07:09,065 INFO [train.py:901] (0/4) Epoch 13, batch 6650, loss[loss=0.2442, simple_loss=0.3156, pruned_loss=0.08642, over 8199.00 frames. ], tot_loss[loss=0.2283, simple_loss=0.3051, pruned_loss=0.07573, over 1612167.88 frames. ], batch size: 23, lr: 5.81e-03, grad_scale: 16.0 +2023-02-06 14:07:16,599 INFO [optim.py:369] (0/4) 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 14:07:43,993 INFO [train.py:901] (0/4) Epoch 13, batch 6700, loss[loss=0.2467, simple_loss=0.3103, pruned_loss=0.09156, over 7971.00 frames. ], tot_loss[loss=0.2282, simple_loss=0.3051, pruned_loss=0.07567, over 1614819.65 frames. ], batch size: 21, lr: 5.81e-03, grad_scale: 16.0 +2023-02-06 14:08:06,460 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 13, batch 6750, loss[loss=0.2443, simple_loss=0.3156, pruned_loss=0.08647, over 8578.00 frames. ], tot_loss[loss=0.2286, simple_loss=0.3056, pruned_loss=0.07577, over 1620210.39 frames. ], batch size: 34, lr: 5.80e-03, grad_scale: 16.0 +2023-02-06 14:08:26,133 INFO [optim.py:369] (0/4) 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 14:08:26,990 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103758.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:08:44,309 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 13, batch 6800, loss[loss=0.2481, simple_loss=0.3194, pruned_loss=0.08838, over 8475.00 frames. ], tot_loss[loss=0.2293, simple_loss=0.3063, pruned_loss=0.07613, over 1618932.05 frames. ], batch size: 29, lr: 5.80e-03, grad_scale: 16.0 +2023-02-06 14:08:58,339 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-06 14:09:17,169 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103843.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 14:09:27,548 INFO [train.py:901] (0/4) Epoch 13, batch 6850, loss[loss=0.3025, simple_loss=0.3621, pruned_loss=0.1215, over 8330.00 frames. ], tot_loss[loss=0.2284, simple_loss=0.3052, pruned_loss=0.0758, over 1613894.47 frames. ], batch size: 49, lr: 5.80e-03, grad_scale: 16.0 +2023-02-06 14:09:33,726 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103856.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:09:34,155 INFO [optim.py:369] (0/4) 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 14:09:40,240 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103865.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:09:44,802 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-06 14:09:57,529 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103890.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:10:00,222 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103894.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:10:02,115 INFO [train.py:901] (0/4) Epoch 13, batch 6900, loss[loss=0.2117, simple_loss=0.2995, pruned_loss=0.06192, over 8259.00 frames. ], tot_loss[loss=0.2296, simple_loss=0.3065, pruned_loss=0.07638, over 1615555.96 frames. ], batch size: 24, lr: 5.80e-03, grad_scale: 8.0 +2023-02-06 14:10:35,837 INFO [train.py:901] (0/4) Epoch 13, batch 6950, loss[loss=0.2115, simple_loss=0.2859, pruned_loss=0.06848, over 8126.00 frames. ], tot_loss[loss=0.2284, simple_loss=0.3055, pruned_loss=0.07564, over 1616943.84 frames. ], batch size: 22, lr: 5.80e-03, grad_scale: 8.0 +2023-02-06 14:10:43,032 INFO [optim.py:369] (0/4) 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 14:10:53,183 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-06 14:11:00,840 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5426, 2.1475, 3.3845, 1.3458, 2.5519, 2.0142, 1.6973, 2.3735], + device='cuda:0'), covar=tensor([0.1768, 0.2226, 0.0748, 0.3957, 0.1609, 0.2802, 0.1886, 0.2157], + device='cuda:0'), in_proj_covar=tensor([0.0496, 0.0537, 0.0531, 0.0588, 0.0620, 0.0559, 0.0483, 0.0613], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 14:11:09,733 INFO [train.py:901] (0/4) Epoch 13, batch 7000, loss[loss=0.247, simple_loss=0.3009, pruned_loss=0.09654, over 7539.00 frames. ], tot_loss[loss=0.2282, simple_loss=0.3051, pruned_loss=0.07561, over 1616593.89 frames. ], batch size: 18, lr: 5.80e-03, grad_scale: 8.0 +2023-02-06 14:11:11,852 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-104000.pt +2023-02-06 14:11:18,193 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104008.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:11:19,230 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 +2023-02-06 14:11:23,537 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104015.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:11:35,614 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2699, 2.3906, 1.8705, 2.9698, 1.4456, 1.7328, 2.1174, 2.3592], + device='cuda:0'), covar=tensor([0.0618, 0.0798, 0.0966, 0.0337, 0.1180, 0.1297, 0.0895, 0.0667], + device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0208, 0.0252, 0.0210, 0.0215, 0.0252, 0.0256, 0.0215], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 14:11:43,224 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.34 vs. limit=5.0 +2023-02-06 14:11:44,816 INFO [train.py:901] (0/4) Epoch 13, batch 7050, loss[loss=0.2142, simple_loss=0.3024, pruned_loss=0.06304, over 8491.00 frames. ], tot_loss[loss=0.2288, simple_loss=0.3056, pruned_loss=0.07598, over 1611921.87 frames. ], batch size: 29, lr: 5.80e-03, grad_scale: 8.0 +2023-02-06 14:11:52,797 INFO [optim.py:369] (0/4) 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 14:11:58,854 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104067.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:12:18,828 INFO [train.py:901] (0/4) Epoch 13, batch 7100, loss[loss=0.1861, simple_loss=0.267, pruned_loss=0.05257, over 8236.00 frames. ], tot_loss[loss=0.2273, simple_loss=0.3043, pruned_loss=0.0752, over 1610641.89 frames. ], batch size: 22, lr: 5.80e-03, grad_scale: 8.0 +2023-02-06 14:12:21,096 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=104099.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 14:12:22,944 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=104102.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:12:37,505 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=104124.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 14:12:39,421 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=104127.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:12:45,744 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4820, 1.4358, 1.7885, 1.4133, 1.1238, 1.8129, 0.1375, 1.0929], + device='cuda:0'), covar=tensor([0.2382, 0.1461, 0.0514, 0.1214, 0.3391, 0.0440, 0.2784, 0.1607], + device='cuda:0'), in_proj_covar=tensor([0.0172, 0.0175, 0.0107, 0.0220, 0.0255, 0.0110, 0.0164, 0.0171], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 14:12:52,701 INFO [train.py:901] (0/4) Epoch 13, batch 7150, loss[loss=0.2478, simple_loss=0.3272, pruned_loss=0.0842, over 8515.00 frames. ], tot_loss[loss=0.2278, simple_loss=0.3044, pruned_loss=0.07553, over 1611121.62 frames. ], batch size: 39, lr: 5.79e-03, grad_scale: 8.0 +2023-02-06 14:13:00,082 INFO [optim.py:369] (0/4) 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 14:13:03,939 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 +2023-02-06 14:13:25,492 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 +2023-02-06 14:13:27,754 INFO [train.py:901] (0/4) Epoch 13, batch 7200, loss[loss=0.1947, simple_loss=0.2826, pruned_loss=0.05336, over 8481.00 frames. ], tot_loss[loss=0.2271, simple_loss=0.304, pruned_loss=0.07506, over 1613144.76 frames. ], batch size: 29, lr: 5.79e-03, grad_scale: 8.0 +2023-02-06 14:13:42,129 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104217.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:13:55,682 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=104238.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:13:58,494 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104242.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:14:01,687 INFO [train.py:901] (0/4) Epoch 13, batch 7250, loss[loss=0.2356, simple_loss=0.3155, pruned_loss=0.07781, over 8490.00 frames. ], tot_loss[loss=0.2282, simple_loss=0.3054, pruned_loss=0.07546, over 1616474.08 frames. ], batch size: 26, lr: 5.79e-03, grad_scale: 8.0 +2023-02-06 14:14:09,619 INFO [optim.py:369] (0/4) 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 14:14:26,702 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8626, 1.8743, 2.5525, 1.9302, 1.4604, 2.5527, 0.5082, 1.6702], + device='cuda:0'), covar=tensor([0.2316, 0.1616, 0.0424, 0.1587, 0.3154, 0.0416, 0.2664, 0.1364], + device='cuda:0'), in_proj_covar=tensor([0.0171, 0.0175, 0.0106, 0.0217, 0.0253, 0.0110, 0.0163, 0.0168], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 14:14:37,319 INFO [train.py:901] (0/4) Epoch 13, batch 7300, loss[loss=0.2317, simple_loss=0.3192, pruned_loss=0.07207, over 8499.00 frames. ], tot_loss[loss=0.2266, simple_loss=0.3038, pruned_loss=0.07469, over 1615223.68 frames. ], batch size: 26, lr: 5.79e-03, grad_scale: 8.0 +2023-02-06 14:14:53,581 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4274, 1.9523, 3.2347, 1.2937, 2.2872, 1.8433, 1.5706, 2.1980], + device='cuda:0'), covar=tensor([0.1765, 0.2219, 0.0750, 0.3950, 0.1653, 0.3001, 0.1879, 0.2217], + device='cuda:0'), in_proj_covar=tensor([0.0501, 0.0543, 0.0538, 0.0595, 0.0624, 0.0564, 0.0491, 0.0619], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 14:14:54,414 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 +2023-02-06 14:15:04,964 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.9519, 1.5456, 6.0421, 2.1095, 5.4244, 5.0973, 5.5866, 5.4999], + device='cuda:0'), covar=tensor([0.0400, 0.4639, 0.0292, 0.3599, 0.0949, 0.0765, 0.0392, 0.0437], + device='cuda:0'), in_proj_covar=tensor([0.0519, 0.0587, 0.0607, 0.0556, 0.0632, 0.0541, 0.0524, 0.0592], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 14:15:11,548 INFO [train.py:901] (0/4) Epoch 13, batch 7350, loss[loss=0.1972, simple_loss=0.2806, pruned_loss=0.05687, over 8242.00 frames. ], tot_loss[loss=0.2264, simple_loss=0.3035, pruned_loss=0.07461, over 1616485.05 frames. ], batch size: 22, lr: 5.79e-03, grad_scale: 8.0 +2023-02-06 14:15:14,978 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104353.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:15:19,101 INFO [optim.py:369] (0/4) 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 14:15:19,895 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=104359.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:15:33,228 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-06 14:15:46,157 INFO [train.py:901] (0/4) Epoch 13, batch 7400, loss[loss=0.1749, simple_loss=0.2581, pruned_loss=0.0458, over 7699.00 frames. ], tot_loss[loss=0.2269, simple_loss=0.3042, pruned_loss=0.07484, over 1614372.74 frames. ], batch size: 18, lr: 5.79e-03, grad_scale: 8.0 +2023-02-06 14:15:50,822 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6061, 1.9028, 1.5551, 2.6869, 1.2960, 1.2964, 1.8938, 2.0705], + device='cuda:0'), covar=tensor([0.1023, 0.0871, 0.1358, 0.0426, 0.1031, 0.1552, 0.0821, 0.0784], + device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0206, 0.0250, 0.0208, 0.0212, 0.0249, 0.0253, 0.0214], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 14:15:53,254 WARNING [train.py:1067] (0/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-06 14:15:56,726 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=104411.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:16:21,106 INFO [train.py:901] (0/4) Epoch 13, batch 7450, loss[loss=0.2206, simple_loss=0.2931, pruned_loss=0.07402, over 8082.00 frames. ], tot_loss[loss=0.2261, simple_loss=0.3034, pruned_loss=0.07435, over 1612301.30 frames. ], batch size: 21, lr: 5.79e-03, grad_scale: 8.0 +2023-02-06 14:16:29,262 INFO [optim.py:369] (0/4) 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 14:16:33,243 WARNING [train.py:1067] (0/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-06 14:16:35,450 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104467.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:16:39,552 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104474.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:16:43,395 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.1944, 3.1329, 2.8619, 1.6335, 2.8353, 2.8085, 2.8489, 2.6728], + device='cuda:0'), covar=tensor([0.1273, 0.0903, 0.1540, 0.4900, 0.1340, 0.1439, 0.1702, 0.1256], + device='cuda:0'), in_proj_covar=tensor([0.0472, 0.0386, 0.0396, 0.0488, 0.0388, 0.0391, 0.0380, 0.0338], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 14:16:55,895 INFO [train.py:901] (0/4) Epoch 13, batch 7500, loss[loss=0.2159, simple_loss=0.3075, pruned_loss=0.06213, over 8608.00 frames. ], tot_loss[loss=0.2283, simple_loss=0.3054, pruned_loss=0.07561, over 1614497.22 frames. ], batch size: 34, lr: 5.78e-03, grad_scale: 8.0 +2023-02-06 14:16:56,748 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=104498.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:16:56,766 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=104498.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:16:59,408 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4486, 2.6201, 1.6824, 2.0985, 2.1679, 1.5657, 1.9405, 2.0462], + device='cuda:0'), covar=tensor([0.1404, 0.0334, 0.1192, 0.0633, 0.0712, 0.1315, 0.0967, 0.0864], + device='cuda:0'), in_proj_covar=tensor([0.0351, 0.0237, 0.0322, 0.0297, 0.0301, 0.0322, 0.0341, 0.0308], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 14:17:14,573 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=104523.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:17:16,621 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104526.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:17:30,657 INFO [train.py:901] (0/4) Epoch 13, batch 7550, loss[loss=0.2932, simple_loss=0.3556, pruned_loss=0.1154, over 8547.00 frames. ], tot_loss[loss=0.2285, simple_loss=0.3056, pruned_loss=0.07567, over 1614665.95 frames. ], batch size: 31, lr: 5.78e-03, grad_scale: 8.0 +2023-02-06 14:17:37,881 INFO [optim.py:369] (0/4) 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 14:17:58,611 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2074, 1.1986, 2.3243, 1.1725, 2.0380, 2.4775, 2.6300, 2.1007], + device='cuda:0'), covar=tensor([0.1207, 0.1410, 0.0483, 0.2117, 0.0779, 0.0381, 0.0604, 0.0788], + device='cuda:0'), in_proj_covar=tensor([0.0270, 0.0303, 0.0264, 0.0294, 0.0278, 0.0239, 0.0358, 0.0294], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 14:18:05,102 INFO [train.py:901] (0/4) Epoch 13, batch 7600, loss[loss=0.2189, simple_loss=0.2789, pruned_loss=0.0794, over 7701.00 frames. ], tot_loss[loss=0.2299, simple_loss=0.3064, pruned_loss=0.07668, over 1615225.88 frames. ], batch size: 18, lr: 5.78e-03, grad_scale: 8.0 +2023-02-06 14:18:13,228 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=104634.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:18:38,448 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.5271, 2.1391, 2.8791, 2.4740, 2.7492, 2.3558, 2.1531, 2.0146], + device='cuda:0'), covar=tensor([0.3226, 0.3614, 0.1261, 0.2489, 0.1738, 0.2194, 0.1344, 0.3280], + device='cuda:0'), in_proj_covar=tensor([0.0894, 0.0890, 0.0746, 0.0872, 0.0950, 0.0824, 0.0706, 0.0780], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 14:18:40,157 INFO [train.py:901] (0/4) Epoch 13, batch 7650, loss[loss=0.2541, simple_loss=0.312, pruned_loss=0.09811, over 8248.00 frames. ], tot_loss[loss=0.2285, simple_loss=0.3049, pruned_loss=0.07603, over 1610178.96 frames. ], batch size: 22, lr: 5.78e-03, grad_scale: 8.0 +2023-02-06 14:18:47,604 INFO [optim.py:369] (0/4) 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 14:19:14,834 INFO [train.py:901] (0/4) Epoch 13, batch 7700, loss[loss=0.2287, simple_loss=0.311, pruned_loss=0.07321, over 8583.00 frames. ], tot_loss[loss=0.2281, simple_loss=0.3047, pruned_loss=0.07577, over 1613798.70 frames. ], batch size: 31, lr: 5.78e-03, grad_scale: 8.0 +2023-02-06 14:19:33,182 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=104723.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:19:37,769 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-06 14:19:37,976 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 13, batch 7750, loss[loss=0.2391, simple_loss=0.3249, pruned_loss=0.07661, over 8113.00 frames. ], tot_loss[loss=0.2298, simple_loss=0.306, pruned_loss=0.07676, over 1618322.21 frames. ], batch size: 23, lr: 5.78e-03, grad_scale: 8.0 +2023-02-06 14:19:50,603 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=104748.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:19:55,991 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 14:19:58,021 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6740, 1.8876, 1.7140, 2.2922, 1.0237, 1.3999, 1.6380, 1.9343], + device='cuda:0'), covar=tensor([0.0853, 0.0868, 0.1020, 0.0491, 0.1207, 0.1508, 0.0920, 0.0789], + device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0208, 0.0254, 0.0211, 0.0213, 0.0253, 0.0256, 0.0215], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 14:20:04,701 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5971, 2.6402, 1.7864, 2.2992, 2.4265, 1.5824, 2.2252, 2.0779], + device='cuda:0'), covar=tensor([0.1502, 0.0365, 0.1159, 0.0668, 0.0633, 0.1500, 0.0833, 0.1032], + device='cuda:0'), in_proj_covar=tensor([0.0346, 0.0234, 0.0317, 0.0293, 0.0298, 0.0320, 0.0337, 0.0303], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 14:20:14,012 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 13, batch 7800, loss[loss=0.236, simple_loss=0.3103, pruned_loss=0.08084, over 8493.00 frames. ], tot_loss[loss=0.229, simple_loss=0.3055, pruned_loss=0.07622, over 1620483.35 frames. ], batch size: 39, lr: 5.78e-03, grad_scale: 8.0 +2023-02-06 14:20:31,895 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=104807.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:20:57,829 INFO [train.py:901] (0/4) Epoch 13, batch 7850, loss[loss=0.2089, simple_loss=0.2718, pruned_loss=0.07304, over 7798.00 frames. ], tot_loss[loss=0.2288, simple_loss=0.3054, pruned_loss=0.07616, over 1620520.31 frames. ], batch size: 19, lr: 5.77e-03, grad_scale: 8.0 +2023-02-06 14:21:05,230 INFO [optim.py:369] (0/4) 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 14:21:30,901 INFO [train.py:901] (0/4) Epoch 13, batch 7900, loss[loss=0.1698, simple_loss=0.2457, pruned_loss=0.04695, over 7703.00 frames. ], tot_loss[loss=0.2292, simple_loss=0.3058, pruned_loss=0.07627, over 1622007.81 frames. ], batch size: 18, lr: 5.77e-03, grad_scale: 8.0 +2023-02-06 14:21:44,399 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8475, 3.7035, 2.1884, 2.5486, 2.8825, 2.1133, 2.5298, 2.8496], + device='cuda:0'), covar=tensor([0.1509, 0.0304, 0.0968, 0.0811, 0.0623, 0.1227, 0.1107, 0.1053], + device='cuda:0'), in_proj_covar=tensor([0.0348, 0.0236, 0.0318, 0.0295, 0.0298, 0.0322, 0.0339, 0.0306], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 14:22:04,230 INFO [train.py:901] (0/4) Epoch 13, batch 7950, loss[loss=0.2595, simple_loss=0.3276, pruned_loss=0.0957, over 8611.00 frames. ], tot_loss[loss=0.2298, simple_loss=0.3063, pruned_loss=0.07666, over 1620493.22 frames. ], batch size: 34, lr: 5.77e-03, grad_scale: 8.0 +2023-02-06 14:22:11,303 INFO [optim.py:369] (0/4) 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 14:22:37,771 INFO [train.py:901] (0/4) Epoch 13, batch 8000, loss[loss=0.2404, simple_loss=0.3265, pruned_loss=0.07714, over 8368.00 frames. ], tot_loss[loss=0.2287, simple_loss=0.3055, pruned_loss=0.07598, over 1620046.15 frames. ], batch size: 24, lr: 5.77e-03, grad_scale: 8.0 +2023-02-06 14:22:47,555 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 +2023-02-06 14:23:10,573 INFO [train.py:901] (0/4) Epoch 13, batch 8050, loss[loss=0.2024, simple_loss=0.2646, pruned_loss=0.07011, over 7547.00 frames. ], tot_loss[loss=0.2272, simple_loss=0.3036, pruned_loss=0.07537, over 1607945.37 frames. ], batch size: 18, lr: 5.77e-03, grad_scale: 8.0 +2023-02-06 14:23:18,074 INFO [optim.py:369] (0/4) 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 14:23:24,941 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0690, 1.1459, 1.2244, 0.8499, 1.2869, 0.9858, 0.3539, 1.1960], + device='cuda:0'), covar=tensor([0.0314, 0.0253, 0.0178, 0.0300, 0.0232, 0.0521, 0.0517, 0.0180], + device='cuda:0'), in_proj_covar=tensor([0.0402, 0.0341, 0.0294, 0.0398, 0.0328, 0.0488, 0.0365, 0.0368], + device='cuda:0'), out_proj_covar=tensor([1.1230e-04, 9.2778e-05, 8.0021e-05, 1.0895e-04, 9.0265e-05, 1.4430e-04, + 1.0224e-04, 1.0168e-04], device='cuda:0') +2023-02-06 14:23:33,781 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/epoch-13.pt +2023-02-06 14:23:50,217 WARNING [train.py:1067] (0/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-06 14:23:54,152 INFO [train.py:901] (0/4) Epoch 14, batch 0, loss[loss=0.2233, simple_loss=0.316, pruned_loss=0.06524, over 8321.00 frames. ], tot_loss[loss=0.2233, simple_loss=0.316, pruned_loss=0.06524, over 8321.00 frames. ], batch size: 26, lr: 5.56e-03, grad_scale: 8.0 +2023-02-06 14:23:54,152 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 14:24:01,264 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5339, 1.8003, 2.5696, 1.3722, 2.1501, 1.8081, 1.6279, 2.0418], + device='cuda:0'), covar=tensor([0.1479, 0.2431, 0.0617, 0.3624, 0.1394, 0.2635, 0.1849, 0.1860], + device='cuda:0'), in_proj_covar=tensor([0.0499, 0.0545, 0.0539, 0.0600, 0.0623, 0.0566, 0.0493, 0.0619], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 14:24:05,196 INFO [train.py:935] (0/4) Epoch 14, validation: loss=0.184, simple_loss=0.2839, pruned_loss=0.04201, over 944034.00 frames. +2023-02-06 14:24:05,197 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-06 14:24:15,528 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.8158, 1.3694, 5.8807, 2.1613, 5.2002, 4.9175, 5.4304, 5.2946], + device='cuda:0'), covar=tensor([0.0465, 0.5287, 0.0373, 0.3539, 0.0994, 0.0817, 0.0483, 0.0524], + device='cuda:0'), in_proj_covar=tensor([0.0520, 0.0581, 0.0608, 0.0551, 0.0632, 0.0536, 0.0523, 0.0589], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 14:24:21,252 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-06 14:24:38,589 INFO [train.py:901] (0/4) Epoch 14, batch 50, loss[loss=0.2156, simple_loss=0.3017, pruned_loss=0.06472, over 8499.00 frames. ], tot_loss[loss=0.2303, simple_loss=0.3093, pruned_loss=0.07564, over 369645.55 frames. ], batch size: 26, lr: 5.56e-03, grad_scale: 8.0 +2023-02-06 14:24:54,776 WARNING [train.py:1067] (0/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-06 14:24:58,156 INFO [optim.py:369] (0/4) 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 14:25:14,420 INFO [train.py:901] (0/4) Epoch 14, batch 100, loss[loss=0.2438, simple_loss=0.3183, pruned_loss=0.0847, over 8579.00 frames. ], tot_loss[loss=0.2299, simple_loss=0.3087, pruned_loss=0.07556, over 651794.72 frames. ], batch size: 34, lr: 5.56e-03, grad_scale: 8.0 +2023-02-06 14:25:17,793 WARNING [train.py:1067] (0/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-06 14:25:48,649 INFO [train.py:901] (0/4) Epoch 14, batch 150, loss[loss=0.1997, simple_loss=0.2821, pruned_loss=0.05866, over 8140.00 frames. ], tot_loss[loss=0.2261, simple_loss=0.3048, pruned_loss=0.07371, over 866176.60 frames. ], batch size: 22, lr: 5.55e-03, grad_scale: 8.0 +2023-02-06 14:26:08,299 INFO [optim.py:369] (0/4) 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 14:26:23,065 INFO [train.py:901] (0/4) Epoch 14, batch 200, loss[loss=0.1984, simple_loss=0.2725, pruned_loss=0.06219, over 7785.00 frames. ], tot_loss[loss=0.2252, simple_loss=0.3037, pruned_loss=0.07338, over 1030602.58 frames. ], batch size: 19, lr: 5.55e-03, grad_scale: 8.0 +2023-02-06 14:26:46,926 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.70 vs. limit=2.0 +2023-02-06 14:26:58,947 INFO [train.py:901] (0/4) Epoch 14, batch 250, loss[loss=0.215, simple_loss=0.2971, pruned_loss=0.06646, over 8565.00 frames. ], tot_loss[loss=0.2244, simple_loss=0.3033, pruned_loss=0.07278, over 1162010.44 frames. ], batch size: 34, lr: 5.55e-03, grad_scale: 8.0 +2023-02-06 14:27:07,608 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-06 14:27:15,953 WARNING [train.py:1067] (0/4) 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] (0/4) 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 14:27:33,653 INFO [train.py:901] (0/4) Epoch 14, batch 300, loss[loss=0.2398, simple_loss=0.3201, pruned_loss=0.07976, over 8361.00 frames. ], tot_loss[loss=0.2246, simple_loss=0.3031, pruned_loss=0.07301, over 1264426.98 frames. ], batch size: 26, lr: 5.55e-03, grad_scale: 8.0 +2023-02-06 14:27:52,811 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=105406.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:28:09,622 INFO [train.py:901] (0/4) Epoch 14, batch 350, loss[loss=0.2272, simple_loss=0.3152, pruned_loss=0.06962, over 8499.00 frames. ], tot_loss[loss=0.2258, simple_loss=0.3045, pruned_loss=0.0736, over 1341409.10 frames. ], batch size: 26, lr: 5.55e-03, grad_scale: 8.0 +2023-02-06 14:28:28,596 INFO [optim.py:369] (0/4) 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 14:28:43,592 INFO [train.py:901] (0/4) Epoch 14, batch 400, loss[loss=0.2255, simple_loss=0.3122, pruned_loss=0.06935, over 8331.00 frames. ], tot_loss[loss=0.2261, simple_loss=0.3047, pruned_loss=0.07378, over 1401857.60 frames. ], batch size: 25, lr: 5.55e-03, grad_scale: 8.0 +2023-02-06 14:29:00,997 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=105504.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:29:13,238 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=105520.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:29:20,756 INFO [train.py:901] (0/4) Epoch 14, batch 450, loss[loss=0.1954, simple_loss=0.2692, pruned_loss=0.06078, over 7797.00 frames. ], tot_loss[loss=0.2248, simple_loss=0.3032, pruned_loss=0.07319, over 1446167.45 frames. ], batch size: 19, lr: 5.55e-03, grad_scale: 8.0 +2023-02-06 14:29:40,049 INFO [optim.py:369] (0/4) 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 14:29:43,640 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.3841, 4.3781, 3.8554, 2.0294, 3.8131, 3.9930, 3.9115, 3.7546], + device='cuda:0'), covar=tensor([0.0754, 0.0583, 0.1268, 0.4930, 0.1025, 0.0938, 0.1315, 0.0827], + device='cuda:0'), in_proj_covar=tensor([0.0474, 0.0390, 0.0399, 0.0498, 0.0390, 0.0393, 0.0379, 0.0341], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 14:29:55,211 INFO [train.py:901] (0/4) Epoch 14, batch 500, loss[loss=0.2078, simple_loss=0.2874, pruned_loss=0.06413, over 8593.00 frames. ], tot_loss[loss=0.2261, simple_loss=0.3043, pruned_loss=0.07395, over 1489992.24 frames. ], batch size: 31, lr: 5.54e-03, grad_scale: 8.0 +2023-02-06 14:30:09,395 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.8034, 3.7219, 3.3505, 1.7568, 3.2957, 3.4410, 3.4123, 3.1460], + device='cuda:0'), covar=tensor([0.0899, 0.0753, 0.1191, 0.4966, 0.1060, 0.1155, 0.1301, 0.1116], + device='cuda:0'), in_proj_covar=tensor([0.0475, 0.0390, 0.0398, 0.0498, 0.0392, 0.0393, 0.0379, 0.0342], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 14:30:29,387 INFO [train.py:901] (0/4) Epoch 14, batch 550, loss[loss=0.243, simple_loss=0.326, pruned_loss=0.08, over 8243.00 frames. ], tot_loss[loss=0.2258, simple_loss=0.3039, pruned_loss=0.07383, over 1521319.12 frames. ], batch size: 22, lr: 5.54e-03, grad_scale: 8.0 +2023-02-06 14:30:50,301 INFO [optim.py:369] (0/4) 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 14:30:56,611 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.6268, 1.4678, 4.8132, 1.6894, 4.2283, 4.0353, 4.3776, 4.2224], + device='cuda:0'), covar=tensor([0.0584, 0.4529, 0.0449, 0.4007, 0.1137, 0.0868, 0.0576, 0.0614], + device='cuda:0'), in_proj_covar=tensor([0.0519, 0.0584, 0.0614, 0.0556, 0.0634, 0.0539, 0.0530, 0.0595], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 14:31:05,192 INFO [train.py:901] (0/4) Epoch 14, batch 600, loss[loss=0.2364, simple_loss=0.3126, pruned_loss=0.0801, over 8340.00 frames. ], tot_loss[loss=0.2246, simple_loss=0.3027, pruned_loss=0.07324, over 1541337.63 frames. ], batch size: 26, lr: 5.54e-03, grad_scale: 8.0 +2023-02-06 14:31:18,466 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-06 14:31:39,816 INFO [train.py:901] (0/4) Epoch 14, batch 650, loss[loss=0.2228, simple_loss=0.2937, pruned_loss=0.07599, over 7925.00 frames. ], tot_loss[loss=0.2236, simple_loss=0.3018, pruned_loss=0.07268, over 1558963.63 frames. ], batch size: 20, lr: 5.54e-03, grad_scale: 8.0 +2023-02-06 14:31:54,394 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 14:32:15,408 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.8078, 1.4296, 3.9380, 1.3037, 3.4275, 3.2753, 3.5928, 3.4594], + device='cuda:0'), covar=tensor([0.0604, 0.4148, 0.0605, 0.3907, 0.1266, 0.0979, 0.0621, 0.0685], + device='cuda:0'), in_proj_covar=tensor([0.0516, 0.0582, 0.0613, 0.0553, 0.0631, 0.0536, 0.0526, 0.0594], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 14:32:17,353 INFO [train.py:901] (0/4) Epoch 14, batch 700, loss[loss=0.1959, simple_loss=0.2682, pruned_loss=0.06179, over 7781.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.3012, pruned_loss=0.07206, over 1572945.19 frames. ], batch size: 19, lr: 5.54e-03, grad_scale: 8.0 +2023-02-06 14:32:23,118 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.99 vs. limit=5.0 +2023-02-06 14:32:34,069 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.96 vs. limit=5.0 +2023-02-06 14:32:37,864 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=105810.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:32:42,733 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=105817.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:32:51,389 INFO [train.py:901] (0/4) Epoch 14, batch 750, loss[loss=0.2234, simple_loss=0.298, pruned_loss=0.07441, over 7973.00 frames. ], tot_loss[loss=0.2224, simple_loss=0.3008, pruned_loss=0.07205, over 1580209.68 frames. ], batch size: 21, lr: 5.54e-03, grad_scale: 8.0 +2023-02-06 14:33:03,852 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=105848.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:33:06,427 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-06 14:33:11,294 INFO [optim.py:369] (0/4) 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 14:33:15,479 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=105864.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:33:16,075 WARNING [train.py:1067] (0/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-06 14:33:16,240 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=105865.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:33:16,864 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3743, 1.1473, 2.4277, 0.9727, 2.1252, 2.0449, 2.2604, 2.1881], + device='cuda:0'), covar=tensor([0.0688, 0.2963, 0.0973, 0.3206, 0.1122, 0.0920, 0.0653, 0.0703], + device='cuda:0'), in_proj_covar=tensor([0.0520, 0.0587, 0.0618, 0.0559, 0.0637, 0.0542, 0.0531, 0.0599], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 14:33:18,522 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.58 vs. limit=5.0 +2023-02-06 14:33:27,180 INFO [train.py:901] (0/4) Epoch 14, batch 800, loss[loss=0.2213, simple_loss=0.294, pruned_loss=0.07436, over 7968.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.3014, pruned_loss=0.07198, over 1588877.81 frames. ], batch size: 21, lr: 5.54e-03, grad_scale: 16.0 +2023-02-06 14:34:02,175 INFO [train.py:901] (0/4) Epoch 14, batch 850, loss[loss=0.2673, simple_loss=0.3398, pruned_loss=0.09734, over 8190.00 frames. ], tot_loss[loss=0.2225, simple_loss=0.3013, pruned_loss=0.0719, over 1597802.52 frames. ], batch size: 23, lr: 5.54e-03, grad_scale: 16.0 +2023-02-06 14:34:20,963 INFO [optim.py:369] (0/4) 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 14:34:24,581 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=105963.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:34:36,565 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=105979.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:34:37,081 INFO [train.py:901] (0/4) Epoch 14, batch 900, loss[loss=0.2207, simple_loss=0.3052, pruned_loss=0.06815, over 8455.00 frames. ], tot_loss[loss=0.2233, simple_loss=0.3016, pruned_loss=0.07243, over 1601826.29 frames. ], batch size: 25, lr: 5.53e-03, grad_scale: 16.0 +2023-02-06 14:34:52,431 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-106000.pt +2023-02-06 14:35:14,899 INFO [train.py:901] (0/4) Epoch 14, batch 950, loss[loss=0.2113, simple_loss=0.289, pruned_loss=0.06682, over 8243.00 frames. ], tot_loss[loss=0.2222, simple_loss=0.3003, pruned_loss=0.07211, over 1604953.75 frames. ], batch size: 22, lr: 5.53e-03, grad_scale: 16.0 +2023-02-06 14:35:15,114 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6082, 1.8964, 2.2982, 1.2406, 2.2900, 1.3976, 0.6940, 1.8018], + device='cuda:0'), covar=tensor([0.0535, 0.0281, 0.0187, 0.0498, 0.0332, 0.0740, 0.0667, 0.0270], + device='cuda:0'), in_proj_covar=tensor([0.0409, 0.0345, 0.0298, 0.0404, 0.0334, 0.0493, 0.0367, 0.0376], + device='cuda:0'), out_proj_covar=tensor([1.1402e-04, 9.3679e-05, 8.1018e-05, 1.1073e-04, 9.1670e-05, 1.4546e-04, + 1.0263e-04, 1.0365e-04], device='cuda:0') +2023-02-06 14:35:33,976 INFO [optim.py:369] (0/4) 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 14:35:38,932 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-06 14:35:49,305 INFO [train.py:901] (0/4) Epoch 14, batch 1000, loss[loss=0.2422, simple_loss=0.3228, pruned_loss=0.08086, over 8542.00 frames. ], tot_loss[loss=0.2221, simple_loss=0.3, pruned_loss=0.07203, over 1608334.01 frames. ], batch size: 49, lr: 5.53e-03, grad_scale: 16.0 +2023-02-06 14:35:51,334 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 +2023-02-06 14:36:00,596 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106095.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:36:14,280 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-06 14:36:20,062 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106121.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:36:26,949 INFO [train.py:901] (0/4) Epoch 14, batch 1050, loss[loss=0.2346, simple_loss=0.3033, pruned_loss=0.08296, over 8126.00 frames. ], tot_loss[loss=0.2224, simple_loss=0.3004, pruned_loss=0.07221, over 1610072.26 frames. ], batch size: 22, lr: 5.53e-03, grad_scale: 16.0 +2023-02-06 14:36:26,960 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-06 14:36:37,948 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106146.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:36:43,486 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=106154.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:36:46,248 INFO [optim.py:369] (0/4) 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 14:36:48,429 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=106161.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:36:54,066 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.4920, 1.4928, 4.7095, 1.5600, 4.1204, 3.9188, 4.2307, 4.0964], + device='cuda:0'), covar=tensor([0.0522, 0.4414, 0.0488, 0.3937, 0.1056, 0.0927, 0.0499, 0.0615], + device='cuda:0'), in_proj_covar=tensor([0.0521, 0.0583, 0.0615, 0.0553, 0.0631, 0.0540, 0.0526, 0.0595], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 14:37:01,618 INFO [train.py:901] (0/4) Epoch 14, batch 1100, loss[loss=0.2584, simple_loss=0.3295, pruned_loss=0.0937, over 7115.00 frames. ], tot_loss[loss=0.2229, simple_loss=0.3013, pruned_loss=0.07224, over 1613398.49 frames. ], batch size: 71, lr: 5.53e-03, grad_scale: 16.0 +2023-02-06 14:37:29,706 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106219.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:37:35,887 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-06 14:37:38,709 INFO [train.py:901] (0/4) Epoch 14, batch 1150, loss[loss=0.2315, simple_loss=0.2944, pruned_loss=0.08427, over 7645.00 frames. ], tot_loss[loss=0.2245, simple_loss=0.3028, pruned_loss=0.07314, over 1608695.45 frames. ], batch size: 19, lr: 5.53e-03, grad_scale: 16.0 +2023-02-06 14:37:42,399 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106235.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:37:49,098 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106244.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:37:58,393 INFO [optim.py:369] (0/4) 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 14:38:00,017 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106260.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:38:03,150 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 +2023-02-06 14:38:06,180 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106269.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:38:10,867 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 14, batch 1200, loss[loss=0.2505, simple_loss=0.3228, pruned_loss=0.08914, over 8456.00 frames. ], tot_loss[loss=0.2242, simple_loss=0.3025, pruned_loss=0.07291, over 1614146.37 frames. ], batch size: 27, lr: 5.53e-03, grad_scale: 16.0 +2023-02-06 14:38:17,531 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106286.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:38:47,936 INFO [train.py:901] (0/4) Epoch 14, batch 1250, loss[loss=0.2185, simple_loss=0.2898, pruned_loss=0.07361, over 8445.00 frames. ], tot_loss[loss=0.2245, simple_loss=0.3027, pruned_loss=0.0731, over 1614844.54 frames. ], batch size: 27, lr: 5.53e-03, grad_scale: 16.0 +2023-02-06 14:39:00,658 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.54 vs. limit=2.0 +2023-02-06 14:39:05,925 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106354.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:39:08,472 INFO [optim.py:369] (0/4) 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 14:39:24,632 INFO [train.py:901] (0/4) Epoch 14, batch 1300, loss[loss=0.2005, simple_loss=0.2827, pruned_loss=0.05912, over 7807.00 frames. ], tot_loss[loss=0.2243, simple_loss=0.3026, pruned_loss=0.07296, over 1614819.07 frames. ], batch size: 20, lr: 5.52e-03, grad_scale: 16.0 +2023-02-06 14:39:52,937 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5286, 2.6992, 1.9011, 2.2422, 2.2750, 1.5397, 2.2063, 2.2524], + device='cuda:0'), covar=tensor([0.1389, 0.0339, 0.1075, 0.0689, 0.0706, 0.1510, 0.0901, 0.0904], + device='cuda:0'), in_proj_covar=tensor([0.0339, 0.0232, 0.0318, 0.0295, 0.0298, 0.0319, 0.0336, 0.0304], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 14:39:58,995 INFO [train.py:901] (0/4) Epoch 14, batch 1350, loss[loss=0.2381, simple_loss=0.3234, pruned_loss=0.07636, over 8513.00 frames. ], tot_loss[loss=0.2225, simple_loss=0.301, pruned_loss=0.07205, over 1611029.77 frames. ], batch size: 28, lr: 5.52e-03, grad_scale: 16.0 +2023-02-06 14:40:05,431 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=106439.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:40:19,203 INFO [optim.py:369] (0/4) 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 14:40:29,770 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106472.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:40:35,128 INFO [train.py:901] (0/4) Epoch 14, batch 1400, loss[loss=0.1926, simple_loss=0.2811, pruned_loss=0.05202, over 7975.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.2996, pruned_loss=0.0713, over 1611327.82 frames. ], batch size: 21, lr: 5.52e-03, grad_scale: 16.0 +2023-02-06 14:40:50,290 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 +2023-02-06 14:41:07,391 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106525.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:41:10,536 INFO [train.py:901] (0/4) Epoch 14, batch 1450, loss[loss=0.2777, simple_loss=0.3461, pruned_loss=0.1047, over 8501.00 frames. ], tot_loss[loss=0.2214, simple_loss=0.2997, pruned_loss=0.07155, over 1611756.52 frames. ], batch size: 26, lr: 5.52e-03, grad_scale: 16.0 +2023-02-06 14:41:11,253 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-06 14:41:12,177 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106532.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:41:24,665 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106550.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:41:27,426 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106554.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:41:29,524 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106557.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:41:29,960 INFO [optim.py:369] (0/4) 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 14:41:41,376 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5897, 2.6641, 1.7649, 2.1512, 2.2279, 1.4874, 2.0501, 2.1306], + device='cuda:0'), covar=tensor([0.1460, 0.0351, 0.1201, 0.0619, 0.0758, 0.1571, 0.1043, 0.0969], + device='cuda:0'), in_proj_covar=tensor([0.0343, 0.0233, 0.0320, 0.0296, 0.0300, 0.0322, 0.0338, 0.0307], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 14:41:47,579 INFO [train.py:901] (0/4) Epoch 14, batch 1500, loss[loss=0.2693, simple_loss=0.3475, pruned_loss=0.09548, over 8341.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.3009, pruned_loss=0.07222, over 1615957.19 frames. ], batch size: 26, lr: 5.52e-03, grad_scale: 16.0 +2023-02-06 14:42:22,567 INFO [train.py:901] (0/4) Epoch 14, batch 1550, loss[loss=0.1858, simple_loss=0.2684, pruned_loss=0.05154, over 7246.00 frames. ], tot_loss[loss=0.2234, simple_loss=0.3016, pruned_loss=0.07259, over 1615143.75 frames. ], batch size: 16, lr: 5.52e-03, grad_scale: 16.0 +2023-02-06 14:42:22,639 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=106630.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:42:41,331 INFO [optim.py:369] (0/4) 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 14:42:56,718 INFO [train.py:901] (0/4) Epoch 14, batch 1600, loss[loss=0.227, simple_loss=0.2964, pruned_loss=0.07883, over 6810.00 frames. ], tot_loss[loss=0.2241, simple_loss=0.3017, pruned_loss=0.07323, over 1612381.72 frames. ], batch size: 15, lr: 5.52e-03, grad_scale: 16.0 +2023-02-06 14:43:10,337 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=106698.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:43:17,988 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 +2023-02-06 14:43:32,334 INFO [train.py:901] (0/4) Epoch 14, batch 1650, loss[loss=0.3238, simple_loss=0.3768, pruned_loss=0.1354, over 8430.00 frames. ], tot_loss[loss=0.2243, simple_loss=0.3019, pruned_loss=0.07333, over 1614908.90 frames. ], batch size: 49, lr: 5.51e-03, grad_scale: 8.0 +2023-02-06 14:43:42,571 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106745.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:43:51,906 INFO [optim.py:369] (0/4) 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 14:44:06,428 INFO [train.py:901] (0/4) Epoch 14, batch 1700, loss[loss=0.2463, simple_loss=0.3199, pruned_loss=0.08632, over 8432.00 frames. ], tot_loss[loss=0.2239, simple_loss=0.3022, pruned_loss=0.07277, over 1618406.68 frames. ], batch size: 49, lr: 5.51e-03, grad_scale: 8.0 +2023-02-06 14:44:28,275 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106810.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:44:31,576 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106813.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:44:33,571 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=106816.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:44:37,203 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0232, 1.4020, 1.6558, 1.3840, 0.9985, 1.4441, 1.8388, 1.6325], + device='cuda:0'), covar=tensor([0.0498, 0.1306, 0.1689, 0.1439, 0.0616, 0.1488, 0.0685, 0.0587], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0152, 0.0191, 0.0158, 0.0101, 0.0162, 0.0114, 0.0138], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 14:44:43,449 INFO [train.py:901] (0/4) Epoch 14, batch 1750, loss[loss=0.2358, simple_loss=0.3142, pruned_loss=0.07874, over 8460.00 frames. ], tot_loss[loss=0.2245, simple_loss=0.303, pruned_loss=0.07303, over 1622489.83 frames. ], batch size: 27, lr: 5.51e-03, grad_scale: 8.0 +2023-02-06 14:44:47,871 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106835.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:45:04,130 INFO [optim.py:369] (0/4) 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 14:45:18,445 INFO [train.py:901] (0/4) Epoch 14, batch 1800, loss[loss=0.1851, simple_loss=0.2653, pruned_loss=0.05243, over 7654.00 frames. ], tot_loss[loss=0.224, simple_loss=0.3022, pruned_loss=0.07287, over 1620411.05 frames. ], batch size: 19, lr: 5.51e-03, grad_scale: 8.0 +2023-02-06 14:45:44,916 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 +2023-02-06 14:45:54,589 INFO [train.py:901] (0/4) Epoch 14, batch 1850, loss[loss=0.2219, simple_loss=0.2878, pruned_loss=0.07799, over 7806.00 frames. ], tot_loss[loss=0.2247, simple_loss=0.3027, pruned_loss=0.07333, over 1615574.58 frames. ], batch size: 19, lr: 5.51e-03, grad_scale: 4.0 +2023-02-06 14:45:55,508 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106931.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:46:00,465 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.59 vs. limit=2.0 +2023-02-06 14:46:16,026 INFO [optim.py:369] (0/4) 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 14:46:17,533 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2451, 1.7970, 1.2972, 2.6157, 1.1914, 1.0894, 1.9085, 1.9316], + device='cuda:0'), covar=tensor([0.1743, 0.1256, 0.2327, 0.0517, 0.1425, 0.2313, 0.0954, 0.0928], + device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0206, 0.0252, 0.0211, 0.0214, 0.0252, 0.0255, 0.0214], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 14:46:23,896 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.60 vs. limit=5.0 +2023-02-06 14:46:29,528 INFO [train.py:901] (0/4) Epoch 14, batch 1900, loss[loss=0.2395, simple_loss=0.3128, pruned_loss=0.08313, over 7969.00 frames. ], tot_loss[loss=0.2245, simple_loss=0.3028, pruned_loss=0.07304, over 1613121.05 frames. ], batch size: 21, lr: 5.51e-03, grad_scale: 4.0 +2023-02-06 14:46:43,884 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107001.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:46:47,084 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-06 14:47:00,490 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-06 14:47:01,253 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 14, batch 1950, loss[loss=0.238, simple_loss=0.3201, pruned_loss=0.07798, over 8518.00 frames. ], tot_loss[loss=0.2254, simple_loss=0.3033, pruned_loss=0.0738, over 1616725.14 frames. ], batch size: 29, lr: 5.51e-03, grad_scale: 4.0 +2023-02-06 14:47:09,964 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.2134, 2.7786, 3.6954, 2.2780, 1.8314, 3.7559, 0.8140, 2.2279], + device='cuda:0'), covar=tensor([0.1597, 0.1753, 0.0305, 0.2034, 0.3642, 0.0249, 0.2846, 0.1855], + device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0177, 0.0107, 0.0218, 0.0260, 0.0112, 0.0163, 0.0175], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 14:47:19,874 WARNING [train.py:1067] (0/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-06 14:47:26,064 INFO [optim.py:369] (0/4) 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 14:47:31,995 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107069.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:47:39,098 INFO [train.py:901] (0/4) Epoch 14, batch 2000, loss[loss=0.1906, simple_loss=0.274, pruned_loss=0.05359, over 7808.00 frames. ], tot_loss[loss=0.223, simple_loss=0.3011, pruned_loss=0.0724, over 1613692.88 frames. ], batch size: 19, lr: 5.51e-03, grad_scale: 8.0 +2023-02-06 14:47:48,646 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107094.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:48:12,595 INFO [train.py:901] (0/4) Epoch 14, batch 2050, loss[loss=0.1906, simple_loss=0.2665, pruned_loss=0.05731, over 7801.00 frames. ], tot_loss[loss=0.2214, simple_loss=0.3, pruned_loss=0.0714, over 1616265.22 frames. ], batch size: 20, lr: 5.50e-03, grad_scale: 8.0 +2023-02-06 14:48:32,533 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=107158.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 14:48:34,367 INFO [optim.py:369] (0/4) 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 14:48:42,106 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=107170.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:48:49,407 INFO [train.py:901] (0/4) Epoch 14, batch 2100, loss[loss=0.3074, simple_loss=0.3746, pruned_loss=0.1201, over 7042.00 frames. ], tot_loss[loss=0.2229, simple_loss=0.3013, pruned_loss=0.07225, over 1615978.82 frames. ], batch size: 73, lr: 5.50e-03, grad_scale: 8.0 +2023-02-06 14:48:49,589 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1767, 1.4906, 1.7250, 1.3470, 0.8253, 1.5147, 1.8131, 1.4666], + device='cuda:0'), covar=tensor([0.0465, 0.1255, 0.1707, 0.1421, 0.0623, 0.1473, 0.0645, 0.0665], + device='cuda:0'), in_proj_covar=tensor([0.0098, 0.0152, 0.0190, 0.0157, 0.0101, 0.0161, 0.0113, 0.0137], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 14:48:54,361 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107187.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:49:11,184 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107212.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:49:23,203 INFO [train.py:901] (0/4) Epoch 14, batch 2150, loss[loss=0.2567, simple_loss=0.3257, pruned_loss=0.0939, over 8498.00 frames. ], tot_loss[loss=0.2243, simple_loss=0.3022, pruned_loss=0.07318, over 1616120.53 frames. ], batch size: 49, lr: 5.50e-03, grad_scale: 8.0 +2023-02-06 14:49:44,430 INFO [optim.py:369] (0/4) 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 14:49:58,901 INFO [train.py:901] (0/4) Epoch 14, batch 2200, loss[loss=0.2355, simple_loss=0.3167, pruned_loss=0.07715, over 8509.00 frames. ], tot_loss[loss=0.2256, simple_loss=0.303, pruned_loss=0.07408, over 1612324.63 frames. ], batch size: 26, lr: 5.50e-03, grad_scale: 8.0 +2023-02-06 14:50:09,035 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 +2023-02-06 14:50:22,066 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 +2023-02-06 14:50:28,934 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 +2023-02-06 14:50:34,528 INFO [train.py:901] (0/4) Epoch 14, batch 2250, loss[loss=0.2362, simple_loss=0.3062, pruned_loss=0.0831, over 8145.00 frames. ], tot_loss[loss=0.2255, simple_loss=0.3025, pruned_loss=0.07423, over 1610840.25 frames. ], batch size: 22, lr: 5.50e-03, grad_scale: 8.0 +2023-02-06 14:50:54,552 INFO [optim.py:369] (0/4) 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 14:51:08,890 INFO [train.py:901] (0/4) Epoch 14, batch 2300, loss[loss=0.2412, simple_loss=0.3139, pruned_loss=0.08424, over 8192.00 frames. ], tot_loss[loss=0.2257, simple_loss=0.3028, pruned_loss=0.07425, over 1616234.90 frames. ], batch size: 23, lr: 5.50e-03, grad_scale: 8.0 +2023-02-06 14:51:11,021 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0847, 2.3256, 1.8574, 2.7803, 1.2693, 1.5830, 1.9798, 2.2728], + device='cuda:0'), covar=tensor([0.0659, 0.0739, 0.0995, 0.0343, 0.1212, 0.1400, 0.0858, 0.0799], + device='cuda:0'), in_proj_covar=tensor([0.0230, 0.0205, 0.0248, 0.0211, 0.0213, 0.0252, 0.0253, 0.0214], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 14:51:44,734 INFO [train.py:901] (0/4) Epoch 14, batch 2350, loss[loss=0.2108, simple_loss=0.2747, pruned_loss=0.07345, over 7423.00 frames. ], tot_loss[loss=0.2256, simple_loss=0.3026, pruned_loss=0.07431, over 1614348.12 frames. ], batch size: 17, lr: 5.50e-03, grad_scale: 8.0 +2023-02-06 14:52:04,937 INFO [optim.py:369] (0/4) 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 14:52:18,380 INFO [train.py:901] (0/4) Epoch 14, batch 2400, loss[loss=0.2172, simple_loss=0.3034, pruned_loss=0.06553, over 8292.00 frames. ], tot_loss[loss=0.2253, simple_loss=0.3025, pruned_loss=0.07409, over 1617671.37 frames. ], batch size: 23, lr: 5.50e-03, grad_scale: 8.0 +2023-02-06 14:52:34,427 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=107502.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 14:52:43,461 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=107514.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:52:53,756 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=107528.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:52:55,004 INFO [train.py:901] (0/4) Epoch 14, batch 2450, loss[loss=0.209, simple_loss=0.2691, pruned_loss=0.07444, over 7799.00 frames. ], tot_loss[loss=0.2238, simple_loss=0.3013, pruned_loss=0.07316, over 1620250.74 frames. ], batch size: 19, lr: 5.49e-03, grad_scale: 8.0 +2023-02-06 14:53:08,642 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.7007, 2.3248, 3.1810, 1.9528, 1.6626, 3.1623, 0.6725, 1.9221], + device='cuda:0'), covar=tensor([0.2081, 0.1342, 0.0328, 0.2156, 0.3647, 0.0406, 0.2956, 0.1938], + device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0177, 0.0108, 0.0219, 0.0261, 0.0111, 0.0162, 0.0175], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 14:53:16,529 INFO [optim.py:369] (0/4) 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 14:53:29,784 INFO [train.py:901] (0/4) Epoch 14, batch 2500, loss[loss=0.262, simple_loss=0.338, pruned_loss=0.09301, over 8025.00 frames. ], tot_loss[loss=0.2236, simple_loss=0.3015, pruned_loss=0.07292, over 1612874.44 frames. ], batch size: 22, lr: 5.49e-03, grad_scale: 8.0 +2023-02-06 14:53:30,963 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 +2023-02-06 14:53:32,367 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.33 vs. limit=5.0 +2023-02-06 14:53:48,274 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8079, 1.8118, 2.3789, 1.5296, 1.1891, 2.4287, 0.3643, 1.4184], + device='cuda:0'), covar=tensor([0.2256, 0.1748, 0.0443, 0.2236, 0.3991, 0.0478, 0.2786, 0.2070], + device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0177, 0.0108, 0.0218, 0.0261, 0.0112, 0.0162, 0.0174], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 14:53:55,573 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=107617.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 14:54:03,464 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=107629.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:54:03,973 INFO [train.py:901] (0/4) Epoch 14, batch 2550, loss[loss=0.2003, simple_loss=0.2853, pruned_loss=0.05771, over 8018.00 frames. ], tot_loss[loss=0.2246, simple_loss=0.3022, pruned_loss=0.07346, over 1612459.75 frames. ], batch size: 22, lr: 5.49e-03, grad_scale: 8.0 +2023-02-06 14:54:13,571 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1944, 2.4076, 2.0307, 2.9082, 1.3697, 1.6880, 2.1124, 2.4839], + device='cuda:0'), covar=tensor([0.0638, 0.0775, 0.0858, 0.0355, 0.1185, 0.1317, 0.0808, 0.0753], + device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0205, 0.0251, 0.0212, 0.0216, 0.0253, 0.0255, 0.0216], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 14:54:26,276 INFO [optim.py:369] (0/4) 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 14:54:37,764 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 14, batch 2600, loss[loss=0.2045, simple_loss=0.2834, pruned_loss=0.06283, over 7923.00 frames. ], tot_loss[loss=0.2235, simple_loss=0.3014, pruned_loss=0.07283, over 1615781.22 frames. ], batch size: 20, lr: 5.49e-03, grad_scale: 8.0 +2023-02-06 14:54:39,762 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3890, 1.4619, 1.4352, 1.7782, 0.6429, 1.2545, 1.2192, 1.4858], + device='cuda:0'), covar=tensor([0.0831, 0.0802, 0.1013, 0.0596, 0.1335, 0.1490, 0.0868, 0.0759], + device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0206, 0.0252, 0.0213, 0.0217, 0.0253, 0.0255, 0.0216], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 14:54:58,281 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.8547, 5.9688, 5.0224, 2.5803, 5.1220, 5.5792, 5.4101, 5.2844], + device='cuda:0'), covar=tensor([0.0448, 0.0358, 0.0839, 0.4020, 0.0614, 0.0512, 0.0999, 0.0473], + device='cuda:0'), in_proj_covar=tensor([0.0487, 0.0401, 0.0408, 0.0501, 0.0401, 0.0402, 0.0388, 0.0349], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 14:55:00,367 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.1767, 4.1444, 3.7618, 1.9789, 3.6677, 3.7119, 3.7739, 3.5299], + device='cuda:0'), covar=tensor([0.0773, 0.0571, 0.1054, 0.4510, 0.0846, 0.0996, 0.1261, 0.0809], + device='cuda:0'), in_proj_covar=tensor([0.0487, 0.0401, 0.0408, 0.0501, 0.0401, 0.0403, 0.0388, 0.0349], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 14:55:12,836 INFO [train.py:901] (0/4) Epoch 14, batch 2650, loss[loss=0.2442, simple_loss=0.3126, pruned_loss=0.08795, over 8453.00 frames. ], tot_loss[loss=0.2232, simple_loss=0.3012, pruned_loss=0.07263, over 1612780.23 frames. ], batch size: 27, lr: 5.49e-03, grad_scale: 8.0 +2023-02-06 14:55:30,855 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=107755.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:55:34,865 INFO [optim.py:369] (0/4) 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 14:55:49,934 INFO [train.py:901] (0/4) Epoch 14, batch 2700, loss[loss=0.3001, simple_loss=0.3497, pruned_loss=0.1252, over 6801.00 frames. ], tot_loss[loss=0.2259, simple_loss=0.3035, pruned_loss=0.07411, over 1614082.77 frames. ], batch size: 72, lr: 5.49e-03, grad_scale: 8.0 +2023-02-06 14:55:54,376 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5985, 2.2562, 4.0296, 1.3228, 2.8172, 2.1032, 1.7667, 2.5843], + device='cuda:0'), covar=tensor([0.2061, 0.2619, 0.0681, 0.4613, 0.1895, 0.3298, 0.2258, 0.2778], + device='cuda:0'), in_proj_covar=tensor([0.0492, 0.0541, 0.0529, 0.0585, 0.0613, 0.0553, 0.0484, 0.0609], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 14:55:58,673 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 +2023-02-06 14:56:02,662 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 +2023-02-06 14:56:23,702 INFO [train.py:901] (0/4) Epoch 14, batch 2750, loss[loss=0.2172, simple_loss=0.2847, pruned_loss=0.07486, over 7700.00 frames. ], tot_loss[loss=0.2246, simple_loss=0.3026, pruned_loss=0.0733, over 1616043.40 frames. ], batch size: 18, lr: 5.49e-03, grad_scale: 8.0 +2023-02-06 14:56:44,700 INFO [optim.py:369] (0/4) 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 14:56:53,308 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=107872.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:56:54,136 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107873.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 14:56:59,459 INFO [train.py:901] (0/4) Epoch 14, batch 2800, loss[loss=0.2202, simple_loss=0.3029, pruned_loss=0.06874, over 8231.00 frames. ], tot_loss[loss=0.2237, simple_loss=0.3021, pruned_loss=0.07267, over 1616693.65 frames. ], batch size: 22, lr: 5.49e-03, grad_scale: 8.0 +2023-02-06 14:57:03,876 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107885.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:57:12,682 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107898.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 14:57:20,759 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107910.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:57:33,816 INFO [train.py:901] (0/4) Epoch 14, batch 2850, loss[loss=0.2603, simple_loss=0.343, pruned_loss=0.08882, over 8357.00 frames. ], tot_loss[loss=0.2259, simple_loss=0.3038, pruned_loss=0.07405, over 1616301.86 frames. ], batch size: 24, lr: 5.48e-03, grad_scale: 8.0 +2023-02-06 14:57:36,750 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([0.8648, 1.6631, 1.3458, 1.5114, 1.3242, 1.1576, 1.1735, 1.2734], + device='cuda:0'), covar=tensor([0.1107, 0.0432, 0.1146, 0.0528, 0.0703, 0.1338, 0.0902, 0.0760], + device='cuda:0'), in_proj_covar=tensor([0.0341, 0.0229, 0.0315, 0.0294, 0.0294, 0.0318, 0.0337, 0.0305], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 14:57:54,109 INFO [optim.py:369] (0/4) 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 14:58:08,144 INFO [train.py:901] (0/4) Epoch 14, batch 2900, loss[loss=0.258, simple_loss=0.3206, pruned_loss=0.09775, over 8241.00 frames. ], tot_loss[loss=0.2255, simple_loss=0.3032, pruned_loss=0.07389, over 1616243.97 frames. ], batch size: 22, lr: 5.48e-03, grad_scale: 8.0 +2023-02-06 14:58:12,777 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=107987.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:58:23,496 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-108000.pt +2023-02-06 14:58:27,870 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-06 14:58:38,625 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=108021.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 14:58:40,705 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.1715, 1.0098, 1.2189, 1.0754, 0.8542, 1.2660, 0.0758, 0.8812], + device='cuda:0'), covar=tensor([0.2112, 0.1823, 0.0583, 0.1071, 0.3513, 0.0567, 0.2844, 0.1572], + device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0178, 0.0108, 0.0219, 0.0261, 0.0112, 0.0163, 0.0175], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 14:58:43,495 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6954, 2.2792, 4.3524, 1.4936, 2.8687, 2.3141, 1.7377, 2.6792], + device='cuda:0'), covar=tensor([0.1711, 0.2316, 0.0670, 0.3976, 0.1722, 0.2782, 0.1935, 0.2474], + device='cuda:0'), in_proj_covar=tensor([0.0497, 0.0546, 0.0538, 0.0590, 0.0620, 0.0557, 0.0489, 0.0616], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 14:58:44,582 INFO [train.py:901] (0/4) Epoch 14, batch 2950, loss[loss=0.2674, simple_loss=0.33, pruned_loss=0.1024, over 8447.00 frames. ], tot_loss[loss=0.2265, simple_loss=0.3041, pruned_loss=0.0744, over 1619669.98 frames. ], batch size: 27, lr: 5.48e-03, grad_scale: 8.0 +2023-02-06 14:58:48,143 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6900, 1.9494, 2.1101, 1.2515, 2.1831, 1.5424, 0.5487, 1.8342], + device='cuda:0'), covar=tensor([0.0409, 0.0256, 0.0206, 0.0391, 0.0286, 0.0667, 0.0620, 0.0199], + device='cuda:0'), in_proj_covar=tensor([0.0416, 0.0354, 0.0309, 0.0410, 0.0341, 0.0500, 0.0376, 0.0382], + device='cuda:0'), out_proj_covar=tensor([1.1604e-04, 9.6232e-05, 8.3977e-05, 1.1160e-04, 9.3310e-05, 1.4720e-04, + 1.0517e-04, 1.0520e-04], device='cuda:0') +2023-02-06 14:58:50,906 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3212, 1.9784, 2.8413, 2.2611, 2.7709, 2.1570, 1.8747, 1.5451], + device='cuda:0'), covar=tensor([0.4483, 0.4336, 0.1423, 0.3077, 0.2112, 0.2534, 0.1692, 0.4432], + device='cuda:0'), in_proj_covar=tensor([0.0903, 0.0900, 0.0743, 0.0870, 0.0954, 0.0830, 0.0712, 0.0781], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 14:59:04,838 INFO [optim.py:369] (0/4) 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 14:59:17,004 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7937, 2.3428, 4.5522, 1.5168, 3.4491, 2.4483, 1.7588, 3.2336], + device='cuda:0'), covar=tensor([0.1614, 0.2379, 0.0667, 0.3777, 0.1233, 0.2612, 0.1898, 0.1970], + device='cuda:0'), in_proj_covar=tensor([0.0497, 0.0547, 0.0539, 0.0591, 0.0618, 0.0558, 0.0489, 0.0617], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 14:59:18,173 INFO [train.py:901] (0/4) Epoch 14, batch 3000, loss[loss=0.2361, simple_loss=0.3147, pruned_loss=0.07871, over 8126.00 frames. ], tot_loss[loss=0.2269, simple_loss=0.3043, pruned_loss=0.07469, over 1617005.41 frames. ], batch size: 22, lr: 5.48e-03, grad_scale: 8.0 +2023-02-06 14:59:18,174 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 14:59:30,512 INFO [train.py:935] (0/4) Epoch 14, validation: loss=0.1827, simple_loss=0.283, pruned_loss=0.04121, over 944034.00 frames. +2023-02-06 14:59:30,513 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-06 14:59:43,702 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=108099.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:00:05,758 INFO [train.py:901] (0/4) Epoch 14, batch 3050, loss[loss=0.2637, simple_loss=0.3401, pruned_loss=0.09361, over 8134.00 frames. ], tot_loss[loss=0.2266, simple_loss=0.3042, pruned_loss=0.07448, over 1615442.68 frames. ], batch size: 22, lr: 5.48e-03, grad_scale: 8.0 +2023-02-06 15:00:10,680 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108136.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:00:17,261 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1493, 1.9047, 2.6770, 2.1710, 2.5075, 2.0739, 1.6981, 1.2382], + device='cuda:0'), covar=tensor([0.4554, 0.4033, 0.1347, 0.2749, 0.2078, 0.2649, 0.1816, 0.4460], + device='cuda:0'), in_proj_covar=tensor([0.0903, 0.0902, 0.0745, 0.0870, 0.0959, 0.0832, 0.0713, 0.0780], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 15:00:28,083 INFO [optim.py:369] (0/4) 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 15:00:41,680 INFO [train.py:901] (0/4) Epoch 14, batch 3100, loss[loss=0.1879, simple_loss=0.2739, pruned_loss=0.0509, over 8139.00 frames. ], tot_loss[loss=0.2247, simple_loss=0.3026, pruned_loss=0.07342, over 1617023.79 frames. ], batch size: 22, lr: 5.48e-03, grad_scale: 8.0 +2023-02-06 15:01:04,696 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108214.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:01:15,953 INFO [train.py:901] (0/4) Epoch 14, batch 3150, loss[loss=0.2104, simple_loss=0.293, pruned_loss=0.06396, over 8622.00 frames. ], tot_loss[loss=0.2233, simple_loss=0.301, pruned_loss=0.07285, over 1613443.28 frames. ], batch size: 34, lr: 5.48e-03, grad_scale: 8.0 +2023-02-06 15:01:24,651 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108243.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:01:37,158 INFO [optim.py:369] (0/4) 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 15:01:42,776 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108268.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:01:51,516 INFO [train.py:901] (0/4) Epoch 14, batch 3200, loss[loss=0.2819, simple_loss=0.3504, pruned_loss=0.1067, over 8198.00 frames. ], tot_loss[loss=0.2238, simple_loss=0.301, pruned_loss=0.07329, over 1611050.67 frames. ], batch size: 23, lr: 5.48e-03, grad_scale: 8.0 +2023-02-06 15:02:13,033 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-06 15:02:25,781 INFO [train.py:901] (0/4) Epoch 14, batch 3250, loss[loss=0.2359, simple_loss=0.3151, pruned_loss=0.07836, over 8505.00 frames. ], tot_loss[loss=0.2236, simple_loss=0.3008, pruned_loss=0.07317, over 1613271.76 frames. ], batch size: 26, lr: 5.47e-03, grad_scale: 8.0 +2023-02-06 15:02:35,719 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108343.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:02:47,021 INFO [optim.py:369] (0/4) 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 15:02:47,296 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6461, 1.6815, 2.0848, 1.4344, 1.0986, 2.0923, 0.2049, 1.2838], + device='cuda:0'), covar=tensor([0.1879, 0.1558, 0.0414, 0.1625, 0.3818, 0.0436, 0.2792, 0.1528], + device='cuda:0'), in_proj_covar=tensor([0.0172, 0.0176, 0.0107, 0.0218, 0.0259, 0.0110, 0.0162, 0.0172], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 15:02:51,341 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8668, 1.5375, 3.1295, 1.3133, 2.1949, 3.3711, 3.4607, 2.8850], + device='cuda:0'), covar=tensor([0.1061, 0.1475, 0.0370, 0.2020, 0.0993, 0.0246, 0.0478, 0.0581], + device='cuda:0'), in_proj_covar=tensor([0.0273, 0.0303, 0.0265, 0.0294, 0.0280, 0.0242, 0.0363, 0.0296], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 15:02:54,347 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.98 vs. limit=5.0 +2023-02-06 15:03:02,203 INFO [train.py:901] (0/4) Epoch 14, batch 3300, loss[loss=0.2542, simple_loss=0.3137, pruned_loss=0.09737, over 7286.00 frames. ], tot_loss[loss=0.225, simple_loss=0.3027, pruned_loss=0.07366, over 1614464.06 frames. ], batch size: 72, lr: 5.47e-03, grad_scale: 8.0 +2023-02-06 15:03:10,829 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.25 vs. limit=5.0 +2023-02-06 15:03:11,254 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108392.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:03:27,913 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108417.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:03:36,476 INFO [train.py:901] (0/4) Epoch 14, batch 3350, loss[loss=0.1983, simple_loss=0.2745, pruned_loss=0.06108, over 7439.00 frames. ], tot_loss[loss=0.2252, simple_loss=0.3027, pruned_loss=0.07381, over 1614033.74 frames. ], batch size: 17, lr: 5.47e-03, grad_scale: 8.0 +2023-02-06 15:03:49,806 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108450.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:03:57,197 INFO [optim.py:369] (0/4) 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 15:04:04,261 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108470.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:04:10,805 INFO [train.py:901] (0/4) Epoch 14, batch 3400, loss[loss=0.2108, simple_loss=0.2822, pruned_loss=0.06967, over 7542.00 frames. ], tot_loss[loss=0.2251, simple_loss=0.3026, pruned_loss=0.0738, over 1612117.64 frames. ], batch size: 18, lr: 5.47e-03, grad_scale: 8.0 +2023-02-06 15:04:22,160 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108495.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:04:46,705 INFO [train.py:901] (0/4) Epoch 14, batch 3450, loss[loss=0.1719, simple_loss=0.2465, pruned_loss=0.04865, over 7698.00 frames. ], tot_loss[loss=0.2248, simple_loss=0.3023, pruned_loss=0.07362, over 1612889.30 frames. ], batch size: 18, lr: 5.47e-03, grad_scale: 8.0 +2023-02-06 15:05:07,907 INFO [optim.py:369] (0/4) 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 15:05:21,992 INFO [train.py:901] (0/4) Epoch 14, batch 3500, loss[loss=0.1886, simple_loss=0.2745, pruned_loss=0.05135, over 8236.00 frames. ], tot_loss[loss=0.2246, simple_loss=0.3021, pruned_loss=0.07356, over 1611363.86 frames. ], batch size: 22, lr: 5.47e-03, grad_scale: 8.0 +2023-02-06 15:05:29,153 WARNING [train.py:1067] (0/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-06 15:05:31,985 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108595.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:05:42,109 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9412, 2.1023, 1.8027, 2.4391, 1.2393, 1.4312, 1.6973, 2.1217], + device='cuda:0'), covar=tensor([0.0674, 0.0799, 0.0910, 0.0521, 0.1213, 0.1533, 0.1020, 0.0718], + device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0206, 0.0250, 0.0212, 0.0214, 0.0249, 0.0255, 0.0213], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 15:05:57,721 INFO [train.py:901] (0/4) Epoch 14, batch 3550, loss[loss=0.2201, simple_loss=0.2943, pruned_loss=0.07292, over 7644.00 frames. ], tot_loss[loss=0.2234, simple_loss=0.3011, pruned_loss=0.07283, over 1609277.27 frames. ], batch size: 19, lr: 5.47e-03, grad_scale: 8.0 +2023-02-06 15:06:17,925 INFO [optim.py:369] (0/4) 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 15:06:31,412 INFO [train.py:901] (0/4) Epoch 14, batch 3600, loss[loss=0.2098, simple_loss=0.2992, pruned_loss=0.06018, over 8367.00 frames. ], tot_loss[loss=0.2235, simple_loss=0.3017, pruned_loss=0.07264, over 1612015.34 frames. ], batch size: 24, lr: 5.47e-03, grad_scale: 8.0 +2023-02-06 15:06:36,122 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=108687.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:07:07,040 INFO [train.py:901] (0/4) Epoch 14, batch 3650, loss[loss=0.1925, simple_loss=0.2818, pruned_loss=0.05159, over 8100.00 frames. ], tot_loss[loss=0.2236, simple_loss=0.3019, pruned_loss=0.07267, over 1611336.46 frames. ], batch size: 23, lr: 5.46e-03, grad_scale: 8.0 +2023-02-06 15:07:27,807 INFO [optim.py:369] (0/4) 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 15:07:30,610 WARNING [train.py:1067] (0/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-06 15:07:41,535 INFO [train.py:901] (0/4) Epoch 14, batch 3700, loss[loss=0.2175, simple_loss=0.2993, pruned_loss=0.06784, over 8448.00 frames. ], tot_loss[loss=0.2241, simple_loss=0.3022, pruned_loss=0.07304, over 1615744.61 frames. ], batch size: 48, lr: 5.46e-03, grad_scale: 8.0 +2023-02-06 15:07:50,766 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=108794.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:07:56,399 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108802.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:08:01,060 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108809.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:08:15,862 INFO [train.py:901] (0/4) Epoch 14, batch 3750, loss[loss=0.2211, simple_loss=0.3017, pruned_loss=0.07024, over 8779.00 frames. ], tot_loss[loss=0.2237, simple_loss=0.3018, pruned_loss=0.07275, over 1617273.90 frames. ], batch size: 30, lr: 5.46e-03, grad_scale: 8.0 +2023-02-06 15:08:37,497 INFO [optim.py:369] (0/4) 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 15:08:51,956 INFO [train.py:901] (0/4) Epoch 14, batch 3800, loss[loss=0.2113, simple_loss=0.2931, pruned_loss=0.06479, over 8762.00 frames. ], tot_loss[loss=0.224, simple_loss=0.302, pruned_loss=0.07296, over 1615548.33 frames. ], batch size: 30, lr: 5.46e-03, grad_scale: 8.0 +2023-02-06 15:08:56,408 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.8700, 2.3559, 3.6945, 2.9354, 3.3551, 2.6124, 2.2897, 1.9044], + device='cuda:0'), covar=tensor([0.4119, 0.4703, 0.1343, 0.2912, 0.1982, 0.2288, 0.1606, 0.4846], + device='cuda:0'), in_proj_covar=tensor([0.0899, 0.0901, 0.0746, 0.0873, 0.0954, 0.0831, 0.0711, 0.0780], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 15:09:12,268 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108909.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:09:26,893 INFO [train.py:901] (0/4) Epoch 14, batch 3850, loss[loss=0.239, simple_loss=0.3142, pruned_loss=0.08192, over 8620.00 frames. ], tot_loss[loss=0.2246, simple_loss=0.3023, pruned_loss=0.07344, over 1615865.97 frames. ], batch size: 31, lr: 5.46e-03, grad_scale: 16.0 +2023-02-06 15:09:33,935 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=108939.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:09:35,948 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-06 15:09:49,068 INFO [optim.py:369] (0/4) 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 15:10:04,295 INFO [train.py:901] (0/4) Epoch 14, batch 3900, loss[loss=0.2372, simple_loss=0.307, pruned_loss=0.08368, over 8624.00 frames. ], tot_loss[loss=0.2233, simple_loss=0.3015, pruned_loss=0.07253, over 1613945.95 frames. ], batch size: 34, lr: 5.46e-03, grad_scale: 16.0 +2023-02-06 15:10:39,039 INFO [train.py:901] (0/4) Epoch 14, batch 3950, loss[loss=0.2319, simple_loss=0.3151, pruned_loss=0.07431, over 8362.00 frames. ], tot_loss[loss=0.2253, simple_loss=0.3027, pruned_loss=0.07392, over 1616316.70 frames. ], batch size: 24, lr: 5.46e-03, grad_scale: 16.0 +2023-02-06 15:10:56,321 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109054.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:10:56,961 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109055.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 15:10:59,085 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109058.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:11:00,259 INFO [optim.py:369] (0/4) 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 15:11:14,761 INFO [train.py:901] (0/4) Epoch 14, batch 4000, loss[loss=0.239, simple_loss=0.3149, pruned_loss=0.08159, over 7969.00 frames. ], tot_loss[loss=0.2253, simple_loss=0.3024, pruned_loss=0.0741, over 1617488.17 frames. ], batch size: 21, lr: 5.46e-03, grad_scale: 16.0 +2023-02-06 15:11:17,688 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109083.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:11:50,557 INFO [train.py:901] (0/4) Epoch 14, batch 4050, loss[loss=0.1912, simple_loss=0.2681, pruned_loss=0.05719, over 7539.00 frames. ], tot_loss[loss=0.2246, simple_loss=0.3024, pruned_loss=0.0734, over 1624266.30 frames. ], batch size: 18, lr: 5.45e-03, grad_scale: 16.0 +2023-02-06 15:12:06,813 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=109153.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:12:11,651 INFO [optim.py:369] (0/4) 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 15:12:16,047 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109165.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:12:19,022 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.71 vs. limit=2.0 +2023-02-06 15:12:26,291 INFO [train.py:901] (0/4) Epoch 14, batch 4100, loss[loss=0.1722, simple_loss=0.2537, pruned_loss=0.04537, over 7204.00 frames. ], tot_loss[loss=0.2241, simple_loss=0.3019, pruned_loss=0.07312, over 1618783.86 frames. ], batch size: 16, lr: 5.45e-03, grad_scale: 16.0 +2023-02-06 15:12:34,006 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109190.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:12:49,793 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109212.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:13:02,503 INFO [train.py:901] (0/4) Epoch 14, batch 4150, loss[loss=0.2143, simple_loss=0.3047, pruned_loss=0.0619, over 8101.00 frames. ], tot_loss[loss=0.2247, simple_loss=0.3025, pruned_loss=0.07345, over 1617427.58 frames. ], batch size: 23, lr: 5.45e-03, grad_scale: 8.0 +2023-02-06 15:13:06,406 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-02-06 15:13:12,246 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.6627, 2.0393, 5.8013, 2.4281, 5.1904, 4.8561, 5.3466, 5.2218], + device='cuda:0'), covar=tensor([0.0447, 0.4001, 0.0327, 0.3044, 0.0879, 0.0761, 0.0420, 0.0445], + device='cuda:0'), in_proj_covar=tensor([0.0531, 0.0600, 0.0619, 0.0561, 0.0634, 0.0545, 0.0535, 0.0597], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 15:13:23,982 INFO [optim.py:369] (0/4) 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 15:13:28,957 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109268.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:13:35,724 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-06 15:13:37,146 INFO [train.py:901] (0/4) Epoch 14, batch 4200, loss[loss=0.1982, simple_loss=0.2808, pruned_loss=0.05782, over 8241.00 frames. ], tot_loss[loss=0.2244, simple_loss=0.302, pruned_loss=0.07341, over 1615410.82 frames. ], batch size: 22, lr: 5.45e-03, grad_scale: 8.0 +2023-02-06 15:13:59,695 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109310.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:14:01,006 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109312.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:14:01,582 WARNING [train.py:1067] (0/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-06 15:14:14,557 INFO [train.py:901] (0/4) Epoch 14, batch 4250, loss[loss=0.1877, simple_loss=0.2758, pruned_loss=0.04983, over 7652.00 frames. ], tot_loss[loss=0.2238, simple_loss=0.3013, pruned_loss=0.07316, over 1612154.95 frames. ], batch size: 19, lr: 5.45e-03, grad_scale: 8.0 +2023-02-06 15:14:18,177 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109335.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:14:18,832 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109336.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:14:35,686 INFO [optim.py:369] (0/4) 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 15:14:48,676 INFO [train.py:901] (0/4) Epoch 14, batch 4300, loss[loss=0.2592, simple_loss=0.3253, pruned_loss=0.09652, over 7821.00 frames. ], tot_loss[loss=0.2235, simple_loss=0.3012, pruned_loss=0.07288, over 1611029.03 frames. ], batch size: 20, lr: 5.45e-03, grad_scale: 8.0 +2023-02-06 15:15:01,898 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=109399.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 15:15:24,501 INFO [train.py:901] (0/4) Epoch 14, batch 4350, loss[loss=0.2337, simple_loss=0.3056, pruned_loss=0.08087, over 7927.00 frames. ], tot_loss[loss=0.2255, simple_loss=0.3027, pruned_loss=0.07415, over 1613142.21 frames. ], batch size: 20, lr: 5.45e-03, grad_scale: 8.0 +2023-02-06 15:15:28,952 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.9200, 2.3310, 3.6260, 2.6311, 3.1960, 2.5933, 2.2067, 1.6855], + device='cuda:0'), covar=tensor([0.3932, 0.4488, 0.1237, 0.2929, 0.2150, 0.2600, 0.1783, 0.4863], + device='cuda:0'), in_proj_covar=tensor([0.0901, 0.0898, 0.0743, 0.0877, 0.0952, 0.0829, 0.0714, 0.0782], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 15:15:34,091 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-06 15:15:47,275 INFO [optim.py:369] (0/4) 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 15:16:00,554 INFO [train.py:901] (0/4) Epoch 14, batch 4400, loss[loss=0.2328, simple_loss=0.3158, pruned_loss=0.0749, over 8288.00 frames. ], tot_loss[loss=0.2244, simple_loss=0.3017, pruned_loss=0.07354, over 1611610.61 frames. ], batch size: 23, lr: 5.45e-03, grad_scale: 8.0 +2023-02-06 15:16:15,770 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-06 15:16:23,486 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.6617, 4.7037, 4.1596, 2.1126, 4.1170, 4.3408, 4.3049, 4.0959], + device='cuda:0'), covar=tensor([0.0655, 0.0495, 0.0987, 0.4781, 0.0794, 0.0944, 0.1224, 0.0730], + device='cuda:0'), in_proj_covar=tensor([0.0473, 0.0390, 0.0396, 0.0490, 0.0391, 0.0394, 0.0380, 0.0341], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 15:16:24,191 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109514.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 15:16:31,903 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109524.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:16:35,904 INFO [train.py:901] (0/4) Epoch 14, batch 4450, loss[loss=0.1996, simple_loss=0.283, pruned_loss=0.05809, over 8194.00 frames. ], tot_loss[loss=0.2242, simple_loss=0.3022, pruned_loss=0.07309, over 1614668.08 frames. ], batch size: 23, lr: 5.44e-03, grad_scale: 8.0 +2023-02-06 15:16:49,302 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109549.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:16:55,376 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=109556.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:16:58,636 INFO [optim.py:369] (0/4) 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 15:17:01,022 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5531, 2.7223, 1.7827, 2.1077, 2.2473, 1.6327, 1.9161, 2.1058], + device='cuda:0'), covar=tensor([0.1346, 0.0312, 0.1010, 0.0646, 0.0584, 0.1196, 0.0980, 0.0851], + device='cuda:0'), in_proj_covar=tensor([0.0345, 0.0231, 0.0319, 0.0294, 0.0296, 0.0324, 0.0342, 0.0307], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 15:17:11,141 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-06 15:17:12,497 INFO [train.py:901] (0/4) Epoch 14, batch 4500, loss[loss=0.237, simple_loss=0.3154, pruned_loss=0.07936, over 8644.00 frames. ], tot_loss[loss=0.2236, simple_loss=0.3016, pruned_loss=0.07284, over 1615292.32 frames. ], batch size: 34, lr: 5.44e-03, grad_scale: 8.0 +2023-02-06 15:17:24,297 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109597.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:17:47,009 INFO [train.py:901] (0/4) Epoch 14, batch 4550, loss[loss=0.1896, simple_loss=0.269, pruned_loss=0.05513, over 7804.00 frames. ], tot_loss[loss=0.2229, simple_loss=0.3007, pruned_loss=0.07256, over 1611877.28 frames. ], batch size: 20, lr: 5.44e-03, grad_scale: 8.0 +2023-02-06 15:18:05,731 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=109656.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:18:09,031 INFO [optim.py:369] (0/4) 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 15:18:16,727 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109671.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:18:21,325 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.6285, 3.5714, 3.2293, 2.1675, 3.1571, 3.2607, 3.3105, 3.0253], + device='cuda:0'), covar=tensor([0.0864, 0.0720, 0.0987, 0.3299, 0.0920, 0.1037, 0.1249, 0.0929], + device='cuda:0'), in_proj_covar=tensor([0.0475, 0.0393, 0.0398, 0.0491, 0.0390, 0.0394, 0.0382, 0.0343], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 15:18:21,367 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109677.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:18:22,082 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0149, 1.6286, 1.2962, 1.5294, 1.3210, 1.1606, 1.2483, 1.2273], + device='cuda:0'), covar=tensor([0.1056, 0.0429, 0.1226, 0.0529, 0.0760, 0.1387, 0.0966, 0.0785], + device='cuda:0'), in_proj_covar=tensor([0.0348, 0.0233, 0.0321, 0.0297, 0.0298, 0.0327, 0.0346, 0.0310], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 15:18:23,249 INFO [train.py:901] (0/4) Epoch 14, batch 4600, loss[loss=0.2233, simple_loss=0.3084, pruned_loss=0.06907, over 8505.00 frames. ], tot_loss[loss=0.2226, simple_loss=0.3003, pruned_loss=0.07242, over 1610326.69 frames. ], batch size: 26, lr: 5.44e-03, grad_scale: 8.0 +2023-02-06 15:18:23,320 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=109680.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:18:48,909 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.64 vs. limit=5.0 +2023-02-06 15:18:56,938 INFO [train.py:901] (0/4) Epoch 14, batch 4650, loss[loss=0.2779, simple_loss=0.321, pruned_loss=0.1174, over 7540.00 frames. ], tot_loss[loss=0.224, simple_loss=0.3016, pruned_loss=0.07318, over 1616121.27 frames. ], batch size: 18, lr: 5.44e-03, grad_scale: 8.0 +2023-02-06 15:19:18,705 INFO [optim.py:369] (0/4) 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 15:19:24,807 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109770.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 15:19:25,410 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109771.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:19:31,303 INFO [train.py:901] (0/4) Epoch 14, batch 4700, loss[loss=0.2703, simple_loss=0.3474, pruned_loss=0.09661, over 8491.00 frames. ], tot_loss[loss=0.2261, simple_loss=0.3037, pruned_loss=0.07419, over 1618644.72 frames. ], batch size: 26, lr: 5.44e-03, grad_scale: 8.0 +2023-02-06 15:19:42,821 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109795.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:19:42,844 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109795.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 15:19:55,433 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.8688, 1.7196, 1.8335, 1.7257, 1.2469, 1.6444, 2.2955, 2.1019], + device='cuda:0'), covar=tensor([0.0414, 0.1163, 0.1593, 0.1243, 0.0547, 0.1401, 0.0579, 0.0546], + device='cuda:0'), in_proj_covar=tensor([0.0098, 0.0154, 0.0191, 0.0158, 0.0102, 0.0163, 0.0115, 0.0138], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 15:20:06,516 INFO [train.py:901] (0/4) Epoch 14, batch 4750, loss[loss=0.2029, simple_loss=0.2933, pruned_loss=0.05622, over 8289.00 frames. ], tot_loss[loss=0.2264, simple_loss=0.3038, pruned_loss=0.07445, over 1616918.70 frames. ], batch size: 23, lr: 5.44e-03, grad_scale: 8.0 +2023-02-06 15:20:10,487 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-06 15:20:12,435 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-06 15:20:26,961 INFO [optim.py:369] (0/4) 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 15:20:39,503 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.16 vs. limit=5.0 +2023-02-06 15:20:40,298 INFO [train.py:901] (0/4) Epoch 14, batch 4800, loss[loss=0.2653, simple_loss=0.3396, pruned_loss=0.09556, over 8190.00 frames. ], tot_loss[loss=0.2273, simple_loss=0.3048, pruned_loss=0.07492, over 1622412.98 frames. ], batch size: 23, lr: 5.44e-03, grad_scale: 8.0 +2023-02-06 15:21:03,196 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-06 15:21:14,258 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109927.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:21:16,044 INFO [train.py:901] (0/4) Epoch 14, batch 4850, loss[loss=0.2121, simple_loss=0.3039, pruned_loss=0.06014, over 8831.00 frames. ], tot_loss[loss=0.2256, simple_loss=0.3036, pruned_loss=0.07383, over 1623962.23 frames. ], batch size: 50, lr: 5.43e-03, grad_scale: 8.0 +2023-02-06 15:21:23,514 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=109941.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:21:31,157 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109952.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:21:37,040 INFO [optim.py:369] (0/4) 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 15:21:49,881 INFO [train.py:901] (0/4) Epoch 14, batch 4900, loss[loss=0.168, simple_loss=0.2455, pruned_loss=0.04527, over 7239.00 frames. ], tot_loss[loss=0.2253, simple_loss=0.303, pruned_loss=0.07385, over 1620540.17 frames. ], batch size: 16, lr: 5.43e-03, grad_scale: 8.0 +2023-02-06 15:22:03,875 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 +2023-02-06 15:22:04,102 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-110000.pt +2023-02-06 15:22:19,432 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=110021.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:22:24,309 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 14, batch 4950, loss[loss=0.222, simple_loss=0.2975, pruned_loss=0.07324, over 8291.00 frames. ], tot_loss[loss=0.2237, simple_loss=0.3019, pruned_loss=0.07274, over 1622256.03 frames. ], batch size: 23, lr: 5.43e-03, grad_scale: 8.0 +2023-02-06 15:22:30,945 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110035.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:22:41,721 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110051.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:22:42,426 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110052.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:22:45,150 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110056.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:22:48,336 INFO [optim.py:369] (0/4) 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 15:22:49,149 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110062.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:22:58,263 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110076.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:23:00,771 INFO [train.py:901] (0/4) Epoch 14, batch 5000, loss[loss=0.2017, simple_loss=0.2869, pruned_loss=0.05819, over 8090.00 frames. ], tot_loss[loss=0.2224, simple_loss=0.3005, pruned_loss=0.07213, over 1616698.68 frames. ], batch size: 21, lr: 5.43e-03, grad_scale: 8.0 +2023-02-06 15:23:04,307 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4299, 1.9714, 3.4164, 1.1890, 2.4352, 1.8833, 1.4695, 2.3096], + device='cuda:0'), covar=tensor([0.1989, 0.2377, 0.0805, 0.4403, 0.1888, 0.3191, 0.2227, 0.2556], + device='cuda:0'), in_proj_covar=tensor([0.0502, 0.0550, 0.0540, 0.0597, 0.0622, 0.0565, 0.0488, 0.0618], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 15:23:10,393 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5336, 1.8752, 2.7801, 1.3941, 1.9427, 1.8853, 1.5975, 1.8999], + device='cuda:0'), covar=tensor([0.1752, 0.2159, 0.0683, 0.3870, 0.1697, 0.2880, 0.1912, 0.2015], + device='cuda:0'), in_proj_covar=tensor([0.0503, 0.0551, 0.0541, 0.0598, 0.0623, 0.0566, 0.0489, 0.0620], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 15:23:33,487 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110128.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:23:34,637 INFO [train.py:901] (0/4) Epoch 14, batch 5050, loss[loss=0.233, simple_loss=0.3056, pruned_loss=0.08023, over 7441.00 frames. ], tot_loss[loss=0.2246, simple_loss=0.3022, pruned_loss=0.0735, over 1615185.34 frames. ], batch size: 17, lr: 5.43e-03, grad_scale: 8.0 +2023-02-06 15:23:36,724 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110133.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:23:38,744 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110136.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:23:43,179 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-06 15:23:47,303 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.7349, 1.2263, 4.8841, 1.6765, 4.3251, 4.1316, 4.4023, 4.2382], + device='cuda:0'), covar=tensor([0.0457, 0.4516, 0.0367, 0.3478, 0.0977, 0.0788, 0.0509, 0.0539], + device='cuda:0'), in_proj_covar=tensor([0.0532, 0.0598, 0.0620, 0.0561, 0.0631, 0.0542, 0.0537, 0.0596], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 15:23:57,272 INFO [optim.py:369] (0/4) 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 15:24:09,945 INFO [train.py:901] (0/4) Epoch 14, batch 5100, loss[loss=0.2404, simple_loss=0.3281, pruned_loss=0.0763, over 8742.00 frames. ], tot_loss[loss=0.2238, simple_loss=0.3017, pruned_loss=0.07292, over 1619521.77 frames. ], batch size: 30, lr: 5.43e-03, grad_scale: 8.0 +2023-02-06 15:24:42,797 INFO [train.py:901] (0/4) Epoch 14, batch 5150, loss[loss=0.1724, simple_loss=0.2572, pruned_loss=0.04379, over 7705.00 frames. ], tot_loss[loss=0.2239, simple_loss=0.3014, pruned_loss=0.07324, over 1609341.17 frames. ], batch size: 18, lr: 5.43e-03, grad_scale: 8.0 +2023-02-06 15:25:05,056 INFO [optim.py:369] (0/4) 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 15:25:19,327 INFO [train.py:901] (0/4) Epoch 14, batch 5200, loss[loss=0.2207, simple_loss=0.2914, pruned_loss=0.07498, over 8101.00 frames. ], tot_loss[loss=0.2248, simple_loss=0.3019, pruned_loss=0.07386, over 1610289.77 frames. ], batch size: 21, lr: 5.43e-03, grad_scale: 8.0 +2023-02-06 15:25:33,781 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110301.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:25:39,485 WARNING [train.py:1067] (0/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-06 15:25:41,071 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110312.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:25:47,137 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110321.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:25:53,186 INFO [train.py:901] (0/4) Epoch 14, batch 5250, loss[loss=0.2098, simple_loss=0.2904, pruned_loss=0.06453, over 7647.00 frames. ], tot_loss[loss=0.226, simple_loss=0.303, pruned_loss=0.07449, over 1609190.62 frames. ], batch size: 19, lr: 5.42e-03, grad_scale: 8.0 +2023-02-06 15:25:57,539 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110336.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:25:58,356 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110337.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:26:08,757 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.7704, 1.4374, 3.9578, 1.4138, 3.4120, 3.3065, 3.5472, 3.4203], + device='cuda:0'), covar=tensor([0.0723, 0.4133, 0.0596, 0.3812, 0.1296, 0.1065, 0.0718, 0.0760], + device='cuda:0'), in_proj_covar=tensor([0.0539, 0.0602, 0.0628, 0.0567, 0.0637, 0.0545, 0.0544, 0.0601], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 15:26:10,243 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.8220, 2.2429, 3.5715, 2.5954, 3.2188, 2.5638, 2.2158, 1.9634], + device='cuda:0'), covar=tensor([0.4241, 0.4818, 0.1390, 0.3004, 0.2201, 0.2547, 0.1804, 0.4695], + device='cuda:0'), in_proj_covar=tensor([0.0898, 0.0900, 0.0744, 0.0877, 0.0952, 0.0830, 0.0714, 0.0782], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 15:26:15,522 INFO [optim.py:369] (0/4) 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 15:26:28,855 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=110379.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:26:29,506 INFO [train.py:901] (0/4) Epoch 14, batch 5300, loss[loss=0.2181, simple_loss=0.2869, pruned_loss=0.07462, over 7787.00 frames. ], tot_loss[loss=0.2256, simple_loss=0.3029, pruned_loss=0.07417, over 1615243.49 frames. ], batch size: 19, lr: 5.42e-03, grad_scale: 8.0 +2023-02-06 15:26:39,505 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110392.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:26:48,908 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=110406.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:26:56,335 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110417.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:27:04,993 INFO [train.py:901] (0/4) Epoch 14, batch 5350, loss[loss=0.2111, simple_loss=0.3024, pruned_loss=0.05993, over 8360.00 frames. ], tot_loss[loss=0.2241, simple_loss=0.3019, pruned_loss=0.07318, over 1617346.44 frames. ], batch size: 24, lr: 5.42e-03, grad_scale: 8.0 +2023-02-06 15:27:25,496 INFO [optim.py:369] (0/4) 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 15:27:32,788 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=110472.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:27:36,872 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=110477.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:27:38,759 INFO [train.py:901] (0/4) Epoch 14, batch 5400, loss[loss=0.2376, simple_loss=0.3098, pruned_loss=0.08268, over 7967.00 frames. ], tot_loss[loss=0.225, simple_loss=0.3029, pruned_loss=0.07355, over 1619266.31 frames. ], batch size: 21, lr: 5.42e-03, grad_scale: 8.0 +2023-02-06 15:27:45,642 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3401, 2.4497, 1.6640, 2.0783, 2.1392, 1.4448, 1.8753, 1.9222], + device='cuda:0'), covar=tensor([0.1420, 0.0365, 0.1180, 0.0550, 0.0624, 0.1453, 0.1035, 0.0825], + device='cuda:0'), in_proj_covar=tensor([0.0347, 0.0232, 0.0319, 0.0294, 0.0295, 0.0325, 0.0339, 0.0308], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 15:27:48,322 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110494.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:28:08,442 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110521.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:28:14,347 INFO [train.py:901] (0/4) Epoch 14, batch 5450, loss[loss=0.2324, simple_loss=0.3248, pruned_loss=0.06999, over 8598.00 frames. ], tot_loss[loss=0.2248, simple_loss=0.3032, pruned_loss=0.07324, over 1620475.35 frames. ], batch size: 31, lr: 5.42e-03, grad_scale: 8.0 +2023-02-06 15:28:30,468 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-06 15:28:34,926 INFO [optim.py:369] (0/4) 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 15:28:46,381 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.6621, 2.9349, 3.4744, 2.1798, 3.4183, 2.4957, 2.0116, 2.6138], + device='cuda:0'), covar=tensor([0.0644, 0.0297, 0.0133, 0.0511, 0.0396, 0.0547, 0.0679, 0.0356], + device='cuda:0'), in_proj_covar=tensor([0.0411, 0.0353, 0.0301, 0.0404, 0.0338, 0.0495, 0.0369, 0.0371], + device='cuda:0'), out_proj_covar=tensor([1.1404e-04, 9.5438e-05, 8.1448e-05, 1.0984e-04, 9.2225e-05, 1.4523e-04, + 1.0269e-04, 1.0161e-04], device='cuda:0') +2023-02-06 15:28:47,533 INFO [train.py:901] (0/4) Epoch 14, batch 5500, loss[loss=0.235, simple_loss=0.3147, pruned_loss=0.07763, over 8132.00 frames. ], tot_loss[loss=0.2245, simple_loss=0.3026, pruned_loss=0.07323, over 1612942.34 frames. ], batch size: 22, lr: 5.42e-03, grad_scale: 8.0 +2023-02-06 15:28:52,315 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110587.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:28:55,800 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110592.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:29:23,308 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110629.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:29:23,834 INFO [train.py:901] (0/4) Epoch 14, batch 5550, loss[loss=0.1782, simple_loss=0.2525, pruned_loss=0.05196, over 7429.00 frames. ], tot_loss[loss=0.2242, simple_loss=0.3014, pruned_loss=0.07344, over 1606435.43 frames. ], batch size: 17, lr: 5.42e-03, grad_scale: 8.0 +2023-02-06 15:29:33,996 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=110645.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:29:44,461 INFO [optim.py:369] (0/4) 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 15:29:44,674 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.7457, 2.2281, 2.3522, 2.2043, 1.4373, 2.3404, 2.5404, 2.3146], + device='cuda:0'), covar=tensor([0.0419, 0.0841, 0.1233, 0.0991, 0.0619, 0.0989, 0.0542, 0.0414], + device='cuda:0'), in_proj_covar=tensor([0.0096, 0.0152, 0.0189, 0.0157, 0.0101, 0.0161, 0.0114, 0.0137], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 15:29:47,074 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=110665.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:29:56,909 INFO [train.py:901] (0/4) Epoch 14, batch 5600, loss[loss=0.2108, simple_loss=0.2818, pruned_loss=0.06996, over 6797.00 frames. ], tot_loss[loss=0.2246, simple_loss=0.3023, pruned_loss=0.0735, over 1611564.52 frames. ], batch size: 15, lr: 5.42e-03, grad_scale: 8.0 +2023-02-06 15:29:56,978 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=110680.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:30:30,942 INFO [train.py:901] (0/4) Epoch 14, batch 5650, loss[loss=0.216, simple_loss=0.2951, pruned_loss=0.06842, over 8295.00 frames. ], tot_loss[loss=0.2236, simple_loss=0.3013, pruned_loss=0.07298, over 1606906.23 frames. ], batch size: 23, lr: 5.41e-03, grad_scale: 8.0 +2023-02-06 15:30:33,073 WARNING [train.py:1067] (0/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-06 15:30:33,295 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6745, 1.6896, 2.2211, 1.4655, 1.1393, 2.2565, 0.2857, 1.3223], + device='cuda:0'), covar=tensor([0.2216, 0.1644, 0.0383, 0.1915, 0.3863, 0.0336, 0.2998, 0.1714], + device='cuda:0'), in_proj_covar=tensor([0.0171, 0.0173, 0.0106, 0.0215, 0.0258, 0.0110, 0.0159, 0.0170], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 15:30:47,233 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110750.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:30:49,875 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110754.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:30:54,018 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 15:31:00,908 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.8903, 2.5854, 3.5613, 1.8569, 1.9261, 3.5828, 0.5622, 1.9560], + device='cuda:0'), covar=tensor([0.2508, 0.1389, 0.0328, 0.2758, 0.3905, 0.0361, 0.3499, 0.2133], + device='cuda:0'), in_proj_covar=tensor([0.0172, 0.0174, 0.0106, 0.0215, 0.0259, 0.0111, 0.0160, 0.0171], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 15:31:04,317 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110775.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:31:05,718 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110777.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:31:07,495 INFO [train.py:901] (0/4) Epoch 14, batch 5700, loss[loss=0.2412, simple_loss=0.3021, pruned_loss=0.09018, over 7966.00 frames. ], tot_loss[loss=0.2216, simple_loss=0.2994, pruned_loss=0.07184, over 1605416.67 frames. ], batch size: 21, lr: 5.41e-03, grad_scale: 8.0 +2023-02-06 15:31:07,674 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110780.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:31:17,910 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110795.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:31:22,885 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110802.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:31:40,955 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-06 15:31:41,631 INFO [train.py:901] (0/4) Epoch 14, batch 5750, loss[loss=0.2527, simple_loss=0.3262, pruned_loss=0.08965, over 8332.00 frames. ], tot_loss[loss=0.2214, simple_loss=0.299, pruned_loss=0.07189, over 1605187.89 frames. ], batch size: 25, lr: 5.41e-03, grad_scale: 8.0 +2023-02-06 15:31:51,523 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110843.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:31:55,000 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110848.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:32:04,409 INFO [optim.py:369] (0/4) 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 15:32:10,160 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110868.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:32:14,063 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110873.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:32:18,685 INFO [train.py:901] (0/4) Epoch 14, batch 5800, loss[loss=0.2253, simple_loss=0.3039, pruned_loss=0.07332, over 8478.00 frames. ], tot_loss[loss=0.2221, simple_loss=0.2995, pruned_loss=0.07233, over 1606041.25 frames. ], batch size: 49, lr: 5.41e-03, grad_scale: 8.0 +2023-02-06 15:32:22,957 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110886.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:32:53,261 INFO [train.py:901] (0/4) Epoch 14, batch 5850, loss[loss=0.2041, simple_loss=0.2815, pruned_loss=0.06339, over 7973.00 frames. ], tot_loss[loss=0.2213, simple_loss=0.2989, pruned_loss=0.07185, over 1605924.71 frames. ], batch size: 21, lr: 5.41e-03, grad_scale: 8.0 +2023-02-06 15:33:14,140 INFO [optim.py:369] (0/4) 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 15:33:22,877 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=110973.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:33:23,754 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3645, 1.9886, 2.8007, 2.3030, 2.7110, 2.2603, 1.9369, 1.4611], + device='cuda:0'), covar=tensor([0.4558, 0.4305, 0.1454, 0.2812, 0.1984, 0.2544, 0.1715, 0.4580], + device='cuda:0'), in_proj_covar=tensor([0.0898, 0.0903, 0.0746, 0.0876, 0.0951, 0.0828, 0.0714, 0.0784], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 15:33:25,733 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110976.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:33:28,980 INFO [train.py:901] (0/4) Epoch 14, batch 5900, loss[loss=0.2376, simple_loss=0.3206, pruned_loss=0.07732, over 8333.00 frames. ], tot_loss[loss=0.2218, simple_loss=0.2996, pruned_loss=0.07202, over 1609071.14 frames. ], batch size: 25, lr: 5.41e-03, grad_scale: 8.0 +2023-02-06 15:33:54,403 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111016.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:34:03,740 INFO [train.py:901] (0/4) Epoch 14, batch 5950, loss[loss=0.2327, simple_loss=0.3164, pruned_loss=0.07453, over 8106.00 frames. ], tot_loss[loss=0.2217, simple_loss=0.2995, pruned_loss=0.07202, over 1607759.00 frames. ], batch size: 23, lr: 5.41e-03, grad_scale: 8.0 +2023-02-06 15:34:07,779 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111036.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:34:11,079 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=111041.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:34:17,712 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111051.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:34:24,226 INFO [optim.py:369] (0/4) 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 15:34:24,447 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=111061.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:34:34,746 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=111076.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:34:38,011 INFO [train.py:901] (0/4) Epoch 14, batch 6000, loss[loss=0.2429, simple_loss=0.319, pruned_loss=0.08342, over 8447.00 frames. ], tot_loss[loss=0.2241, simple_loss=0.3017, pruned_loss=0.07331, over 1612373.05 frames. ], batch size: 29, lr: 5.41e-03, grad_scale: 8.0 +2023-02-06 15:34:38,012 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 15:34:50,549 INFO [train.py:935] (0/4) Epoch 14, validation: loss=0.1818, simple_loss=0.2816, pruned_loss=0.04094, over 944034.00 frames. +2023-02-06 15:34:50,550 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-06 15:34:56,293 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111088.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:35:03,693 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=111098.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:35:27,232 INFO [train.py:901] (0/4) Epoch 14, batch 6050, loss[loss=0.2398, simple_loss=0.3095, pruned_loss=0.08509, over 7935.00 frames. ], tot_loss[loss=0.223, simple_loss=0.3005, pruned_loss=0.07273, over 1609149.34 frames. ], batch size: 20, lr: 5.40e-03, grad_scale: 4.0 +2023-02-06 15:35:49,432 INFO [optim.py:369] (0/4) 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 15:36:01,588 INFO [train.py:901] (0/4) Epoch 14, batch 6100, loss[loss=0.2413, simple_loss=0.3186, pruned_loss=0.08204, over 8347.00 frames. ], tot_loss[loss=0.221, simple_loss=0.2988, pruned_loss=0.07167, over 1610343.70 frames. ], batch size: 26, lr: 5.40e-03, grad_scale: 4.0 +2023-02-06 15:36:03,139 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6262, 1.5622, 2.8319, 1.2928, 2.1722, 3.0425, 3.1382, 2.5838], + device='cuda:0'), covar=tensor([0.1116, 0.1417, 0.0387, 0.2021, 0.0828, 0.0293, 0.0570, 0.0628], + device='cuda:0'), in_proj_covar=tensor([0.0272, 0.0305, 0.0268, 0.0296, 0.0283, 0.0244, 0.0370, 0.0293], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-06 15:36:09,227 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4962, 1.9005, 1.9295, 1.0060, 1.9970, 1.4521, 0.4294, 1.7328], + device='cuda:0'), covar=tensor([0.0426, 0.0274, 0.0209, 0.0436, 0.0321, 0.0663, 0.0655, 0.0230], + device='cuda:0'), in_proj_covar=tensor([0.0415, 0.0356, 0.0303, 0.0406, 0.0342, 0.0496, 0.0371, 0.0380], + device='cuda:0'), out_proj_covar=tensor([1.1529e-04, 9.6312e-05, 8.2072e-05, 1.1021e-04, 9.3345e-05, 1.4527e-04, + 1.0353e-04, 1.0408e-04], device='cuda:0') +2023-02-06 15:36:15,953 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-06 15:36:24,950 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111213.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:36:37,121 INFO [train.py:901] (0/4) Epoch 14, batch 6150, loss[loss=0.1909, simple_loss=0.2666, pruned_loss=0.05764, over 7408.00 frames. ], tot_loss[loss=0.2205, simple_loss=0.2988, pruned_loss=0.07113, over 1608110.97 frames. ], batch size: 17, lr: 5.40e-03, grad_scale: 4.0 +2023-02-06 15:36:37,202 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=111230.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:36:46,706 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111243.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:36:59,323 INFO [optim.py:369] (0/4) 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 15:37:11,852 INFO [train.py:901] (0/4) Epoch 14, batch 6200, loss[loss=0.2139, simple_loss=0.2968, pruned_loss=0.06551, over 8761.00 frames. ], tot_loss[loss=0.2208, simple_loss=0.2989, pruned_loss=0.07141, over 1606959.03 frames. ], batch size: 30, lr: 5.40e-03, grad_scale: 4.0 +2023-02-06 15:37:38,511 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 14, batch 6250, loss[loss=0.2012, simple_loss=0.2888, pruned_loss=0.05673, over 8286.00 frames. ], tot_loss[loss=0.2192, simple_loss=0.2975, pruned_loss=0.07041, over 1607102.50 frames. ], batch size: 23, lr: 5.40e-03, grad_scale: 4.0 +2023-02-06 15:37:55,173 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111344.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:37:55,810 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111345.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:38:08,443 INFO [optim.py:369] (0/4) 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 15:38:13,369 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=111369.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:38:20,509 INFO [train.py:901] (0/4) Epoch 14, batch 6300, loss[loss=0.2177, simple_loss=0.3076, pruned_loss=0.06385, over 8497.00 frames. ], tot_loss[loss=0.22, simple_loss=0.2989, pruned_loss=0.07055, over 1610463.20 frames. ], batch size: 49, lr: 5.40e-03, grad_scale: 4.0 +2023-02-06 15:38:55,242 INFO [train.py:901] (0/4) Epoch 14, batch 6350, loss[loss=0.2386, simple_loss=0.3117, pruned_loss=0.08282, over 8184.00 frames. ], tot_loss[loss=0.2205, simple_loss=0.2993, pruned_loss=0.07085, over 1607737.80 frames. ], batch size: 23, lr: 5.40e-03, grad_scale: 4.0 +2023-02-06 15:38:58,944 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111435.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:39:12,613 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 +2023-02-06 15:39:13,910 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 +2023-02-06 15:39:17,455 INFO [optim.py:369] (0/4) 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 15:39:23,082 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 14, batch 6400, loss[loss=0.1857, simple_loss=0.2577, pruned_loss=0.05682, over 7690.00 frames. ], tot_loss[loss=0.2198, simple_loss=0.2986, pruned_loss=0.07056, over 1605440.32 frames. ], batch size: 18, lr: 5.40e-03, grad_scale: 8.0 +2023-02-06 15:39:32,442 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.5678, 4.5434, 4.1060, 1.9655, 4.0183, 4.2483, 4.1669, 4.0469], + device='cuda:0'), covar=tensor([0.0708, 0.0583, 0.1057, 0.5093, 0.0867, 0.0900, 0.1280, 0.0725], + device='cuda:0'), in_proj_covar=tensor([0.0484, 0.0400, 0.0402, 0.0500, 0.0395, 0.0401, 0.0388, 0.0346], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 15:39:35,198 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111486.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:39:40,683 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=111494.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:39:54,996 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.96 vs. limit=5.0 +2023-02-06 15:40:05,256 INFO [train.py:901] (0/4) Epoch 14, batch 6450, loss[loss=0.2045, simple_loss=0.2677, pruned_loss=0.07064, over 7703.00 frames. ], tot_loss[loss=0.2188, simple_loss=0.2979, pruned_loss=0.0699, over 1608011.32 frames. ], batch size: 18, lr: 5.40e-03, grad_scale: 8.0 +2023-02-06 15:40:26,320 INFO [optim.py:369] (0/4) 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 15:40:37,968 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2257, 1.9416, 2.7435, 2.2665, 2.5596, 2.1594, 1.7834, 1.3579], + device='cuda:0'), covar=tensor([0.4356, 0.4051, 0.1303, 0.2611, 0.1993, 0.2301, 0.1757, 0.4327], + device='cuda:0'), in_proj_covar=tensor([0.0893, 0.0897, 0.0744, 0.0872, 0.0941, 0.0828, 0.0710, 0.0780], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 15:40:39,078 INFO [train.py:901] (0/4) Epoch 14, batch 6500, loss[loss=0.2147, simple_loss=0.2989, pruned_loss=0.06523, over 8629.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.2994, pruned_loss=0.07138, over 1607582.68 frames. ], batch size: 39, lr: 5.39e-03, grad_scale: 8.0 +2023-02-06 15:40:44,378 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=111587.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:40:55,105 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111601.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:41:12,085 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 14, batch 6550, loss[loss=0.2272, simple_loss=0.3041, pruned_loss=0.07514, over 8561.00 frames. ], tot_loss[loss=0.222, simple_loss=0.2998, pruned_loss=0.07205, over 1610681.77 frames. ], batch size: 31, lr: 5.39e-03, grad_scale: 8.0 +2023-02-06 15:41:24,436 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-06 15:41:35,843 INFO [optim.py:369] (0/4) 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 15:41:42,133 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9563, 1.8584, 2.4617, 1.5738, 1.2880, 2.5552, 0.4148, 1.3760], + device='cuda:0'), covar=tensor([0.2006, 0.1450, 0.0319, 0.1793, 0.3287, 0.0353, 0.2946, 0.1708], + device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0177, 0.0109, 0.0217, 0.0260, 0.0112, 0.0164, 0.0174], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 15:41:43,336 WARNING [train.py:1067] (0/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-06 15:41:47,951 INFO [train.py:901] (0/4) Epoch 14, batch 6600, loss[loss=0.2255, simple_loss=0.2977, pruned_loss=0.07661, over 8128.00 frames. ], tot_loss[loss=0.2238, simple_loss=0.3016, pruned_loss=0.07305, over 1611235.80 frames. ], batch size: 22, lr: 5.39e-03, grad_scale: 8.0 +2023-02-06 15:41:55,623 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111691.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:42:03,605 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111702.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:42:14,379 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=111716.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:42:23,469 INFO [train.py:901] (0/4) Epoch 14, batch 6650, loss[loss=0.2063, simple_loss=0.2853, pruned_loss=0.06371, over 7964.00 frames. ], tot_loss[loss=0.2233, simple_loss=0.3004, pruned_loss=0.07308, over 1611460.33 frames. ], batch size: 21, lr: 5.39e-03, grad_scale: 8.0 +2023-02-06 15:42:26,330 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6368, 2.8125, 1.8826, 2.3320, 2.2032, 1.6134, 2.1659, 2.3766], + device='cuda:0'), covar=tensor([0.1653, 0.0377, 0.1146, 0.0677, 0.0710, 0.1398, 0.1075, 0.0957], + device='cuda:0'), in_proj_covar=tensor([0.0350, 0.0235, 0.0326, 0.0297, 0.0301, 0.0329, 0.0344, 0.0311], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 15:42:45,229 INFO [optim.py:369] (0/4) 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 15:42:57,308 INFO [train.py:901] (0/4) Epoch 14, batch 6700, loss[loss=0.2497, simple_loss=0.3206, pruned_loss=0.0894, over 8464.00 frames. ], tot_loss[loss=0.2236, simple_loss=0.3007, pruned_loss=0.07323, over 1616362.43 frames. ], batch size: 25, lr: 5.39e-03, grad_scale: 8.0 +2023-02-06 15:43:16,619 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111809.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 15:43:32,519 INFO [train.py:901] (0/4) Epoch 14, batch 6750, loss[loss=0.2699, simple_loss=0.3383, pruned_loss=0.1008, over 6882.00 frames. ], tot_loss[loss=0.2221, simple_loss=0.3, pruned_loss=0.07204, over 1617820.26 frames. ], batch size: 71, lr: 5.39e-03, grad_scale: 8.0 +2023-02-06 15:43:32,588 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=111830.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:43:54,031 INFO [optim.py:369] (0/4) 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 15:44:02,382 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-06 15:44:03,213 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.8554, 3.7784, 3.4895, 1.6480, 3.4010, 3.3871, 3.4685, 3.1906], + device='cuda:0'), covar=tensor([0.0857, 0.0720, 0.1051, 0.5011, 0.0917, 0.1081, 0.1466, 0.0884], + device='cuda:0'), in_proj_covar=tensor([0.0479, 0.0395, 0.0399, 0.0495, 0.0392, 0.0397, 0.0384, 0.0340], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 15:44:07,192 INFO [train.py:901] (0/4) Epoch 14, batch 6800, loss[loss=0.2288, simple_loss=0.2906, pruned_loss=0.08344, over 7561.00 frames. ], tot_loss[loss=0.2222, simple_loss=0.3004, pruned_loss=0.07198, over 1616093.30 frames. ], batch size: 18, lr: 5.39e-03, grad_scale: 8.0 +2023-02-06 15:44:14,363 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.30 vs. limit=5.0 +2023-02-06 15:44:40,412 INFO [train.py:901] (0/4) Epoch 14, batch 6850, loss[loss=0.2603, simple_loss=0.3457, pruned_loss=0.08743, over 8524.00 frames. ], tot_loss[loss=0.2225, simple_loss=0.3008, pruned_loss=0.07209, over 1616697.75 frames. ], batch size: 28, lr: 5.39e-03, grad_scale: 8.0 +2023-02-06 15:44:51,192 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-06 15:44:52,047 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111945.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:45:01,401 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 15:45:16,599 INFO [train.py:901] (0/4) Epoch 14, batch 6900, loss[loss=0.2279, simple_loss=0.3021, pruned_loss=0.07689, over 8454.00 frames. ], tot_loss[loss=0.2233, simple_loss=0.3014, pruned_loss=0.07261, over 1616400.98 frames. ], batch size: 27, lr: 5.38e-03, grad_scale: 8.0 +2023-02-06 15:45:18,704 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=111983.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:45:23,422 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 +2023-02-06 15:45:29,849 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-112000.pt +2023-02-06 15:45:50,585 INFO [train.py:901] (0/4) Epoch 14, batch 6950, loss[loss=0.2281, simple_loss=0.3087, pruned_loss=0.07369, over 7804.00 frames. ], tot_loss[loss=0.2234, simple_loss=0.3008, pruned_loss=0.07296, over 1613049.06 frames. ], batch size: 20, lr: 5.38e-03, grad_scale: 8.0 +2023-02-06 15:45:58,675 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-06 15:46:02,907 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6859, 2.8806, 1.9238, 2.4031, 2.4210, 1.6599, 2.3636, 2.3356], + device='cuda:0'), covar=tensor([0.1492, 0.0411, 0.1143, 0.0649, 0.0603, 0.1473, 0.0820, 0.0993], + device='cuda:0'), in_proj_covar=tensor([0.0353, 0.0236, 0.0328, 0.0302, 0.0304, 0.0332, 0.0348, 0.0313], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 15:46:13,920 INFO [optim.py:369] (0/4) 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 15:46:25,965 INFO [train.py:901] (0/4) Epoch 14, batch 7000, loss[loss=0.2056, simple_loss=0.2836, pruned_loss=0.06385, over 7554.00 frames. ], tot_loss[loss=0.2233, simple_loss=0.3007, pruned_loss=0.07294, over 1608142.28 frames. ], batch size: 18, lr: 5.38e-03, grad_scale: 8.0 +2023-02-06 15:46:28,772 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112084.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:46:29,104 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 +2023-02-06 15:46:59,913 INFO [train.py:901] (0/4) Epoch 14, batch 7050, loss[loss=0.1905, simple_loss=0.2729, pruned_loss=0.05403, over 8037.00 frames. ], tot_loss[loss=0.224, simple_loss=0.3012, pruned_loss=0.07341, over 1607765.19 frames. ], batch size: 22, lr: 5.38e-03, grad_scale: 8.0 +2023-02-06 15:47:15,899 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=112153.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 15:47:17,937 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.6960, 1.5723, 1.7340, 1.5558, 1.0209, 1.6622, 2.0461, 1.9744], + device='cuda:0'), covar=tensor([0.0476, 0.1264, 0.1761, 0.1430, 0.0626, 0.1493, 0.0685, 0.0573], + device='cuda:0'), in_proj_covar=tensor([0.0096, 0.0151, 0.0188, 0.0154, 0.0100, 0.0160, 0.0114, 0.0136], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0008, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 15:47:21,802 INFO [optim.py:369] (0/4) 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 15:47:30,690 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2149, 2.0085, 3.2286, 1.9847, 2.6886, 3.5621, 3.4999, 3.2298], + device='cuda:0'), covar=tensor([0.0967, 0.1353, 0.0503, 0.1546, 0.1296, 0.0217, 0.0603, 0.0446], + device='cuda:0'), in_proj_covar=tensor([0.0279, 0.0308, 0.0270, 0.0301, 0.0287, 0.0246, 0.0373, 0.0297], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 15:47:35,265 INFO [train.py:901] (0/4) Epoch 14, batch 7100, loss[loss=0.2005, simple_loss=0.2776, pruned_loss=0.06167, over 7921.00 frames. ], tot_loss[loss=0.2247, simple_loss=0.302, pruned_loss=0.07372, over 1605572.40 frames. ], batch size: 20, lr: 5.38e-03, grad_scale: 8.0 +2023-02-06 15:47:50,282 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112201.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:48:07,694 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 14, batch 7150, loss[loss=0.25, simple_loss=0.3194, pruned_loss=0.09027, over 6778.00 frames. ], tot_loss[loss=0.224, simple_loss=0.3014, pruned_loss=0.07327, over 1602999.18 frames. ], batch size: 71, lr: 5.38e-03, grad_scale: 8.0 +2023-02-06 15:48:11,592 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3676, 3.0683, 2.2103, 3.9446, 1.8543, 1.8666, 2.4674, 3.0817], + device='cuda:0'), covar=tensor([0.0708, 0.0730, 0.0941, 0.0220, 0.1027, 0.1373, 0.0946, 0.0790], + device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0206, 0.0253, 0.0214, 0.0216, 0.0252, 0.0258, 0.0215], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 15:48:24,024 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 +2023-02-06 15:48:31,540 INFO [optim.py:369] (0/4) 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 15:48:35,572 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=112268.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 15:48:43,203 INFO [train.py:901] (0/4) Epoch 14, batch 7200, loss[loss=0.1881, simple_loss=0.2749, pruned_loss=0.05062, over 7966.00 frames. ], tot_loss[loss=0.2219, simple_loss=0.2994, pruned_loss=0.07224, over 1595238.45 frames. ], batch size: 21, lr: 5.38e-03, grad_scale: 8.0 +2023-02-06 15:49:18,417 INFO [train.py:901] (0/4) Epoch 14, batch 7250, loss[loss=0.1885, simple_loss=0.2769, pruned_loss=0.04999, over 8189.00 frames. ], tot_loss[loss=0.2215, simple_loss=0.2989, pruned_loss=0.072, over 1601443.75 frames. ], batch size: 23, lr: 5.38e-03, grad_scale: 8.0 +2023-02-06 15:49:31,016 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 +2023-02-06 15:49:39,902 INFO [optim.py:369] (0/4) 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 15:49:52,058 INFO [train.py:901] (0/4) Epoch 14, batch 7300, loss[loss=0.2366, simple_loss=0.3192, pruned_loss=0.07702, over 8689.00 frames. ], tot_loss[loss=0.2225, simple_loss=0.2998, pruned_loss=0.07258, over 1605987.65 frames. ], batch size: 34, lr: 5.37e-03, grad_scale: 8.0 +2023-02-06 15:50:11,919 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112407.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:50:26,698 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 14, batch 7350, loss[loss=0.2161, simple_loss=0.2989, pruned_loss=0.06663, over 8561.00 frames. ], tot_loss[loss=0.2232, simple_loss=0.3005, pruned_loss=0.07296, over 1609628.95 frames. ], batch size: 31, lr: 5.37e-03, grad_scale: 8.0 +2023-02-06 15:50:40,032 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-06 15:50:49,901 INFO [optim.py:369] (0/4) 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 15:50:59,944 WARNING [train.py:1067] (0/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-06 15:51:02,055 INFO [train.py:901] (0/4) Epoch 14, batch 7400, loss[loss=0.2267, simple_loss=0.309, pruned_loss=0.07219, over 8288.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.3006, pruned_loss=0.07241, over 1608889.32 frames. ], batch size: 23, lr: 5.37e-03, grad_scale: 8.0 +2023-02-06 15:51:32,626 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112524.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 15:51:37,190 INFO [train.py:901] (0/4) Epoch 14, batch 7450, loss[loss=0.2327, simple_loss=0.2936, pruned_loss=0.0859, over 7704.00 frames. ], tot_loss[loss=0.223, simple_loss=0.3007, pruned_loss=0.07264, over 1609360.77 frames. ], batch size: 18, lr: 5.37e-03, grad_scale: 8.0 +2023-02-06 15:51:41,803 WARNING [train.py:1067] (0/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-06 15:51:46,949 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=112543.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:51:50,981 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112549.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 15:51:59,931 INFO [optim.py:369] (0/4) 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 15:52:13,215 INFO [train.py:901] (0/4) Epoch 14, batch 7500, loss[loss=0.232, simple_loss=0.3076, pruned_loss=0.07817, over 8579.00 frames. ], tot_loss[loss=0.2231, simple_loss=0.3009, pruned_loss=0.07262, over 1609269.43 frames. ], batch size: 39, lr: 5.37e-03, grad_scale: 8.0 +2023-02-06 15:52:13,360 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112580.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:52:23,781 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112595.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:52:47,276 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1910, 1.8367, 2.6921, 2.2192, 2.6484, 2.0677, 1.7337, 1.3513], + device='cuda:0'), covar=tensor([0.4726, 0.4629, 0.1573, 0.2923, 0.2157, 0.2820, 0.2066, 0.4838], + device='cuda:0'), in_proj_covar=tensor([0.0902, 0.0909, 0.0750, 0.0878, 0.0953, 0.0837, 0.0722, 0.0788], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 15:52:47,704 INFO [train.py:901] (0/4) Epoch 14, batch 7550, loss[loss=0.2554, simple_loss=0.3366, pruned_loss=0.08708, over 8102.00 frames. ], tot_loss[loss=0.2237, simple_loss=0.3015, pruned_loss=0.07292, over 1610540.17 frames. ], batch size: 23, lr: 5.37e-03, grad_scale: 8.0 +2023-02-06 15:52:51,278 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) Epoch 14, batch 7600, loss[loss=0.2152, simple_loss=0.293, pruned_loss=0.06872, over 8036.00 frames. ], tot_loss[loss=0.2237, simple_loss=0.3014, pruned_loss=0.07302, over 1611288.25 frames. ], batch size: 22, lr: 5.37e-03, grad_scale: 8.0 +2023-02-06 15:53:39,793 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.7842, 2.2988, 3.3841, 1.8035, 1.5901, 3.3902, 0.5285, 1.8490], + device='cuda:0'), covar=tensor([0.1934, 0.1543, 0.0293, 0.2413, 0.4056, 0.0336, 0.3360, 0.2195], + device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0178, 0.0110, 0.0218, 0.0259, 0.0113, 0.0163, 0.0176], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 15:53:57,107 INFO [train.py:901] (0/4) Epoch 14, batch 7650, loss[loss=0.2059, simple_loss=0.2664, pruned_loss=0.07272, over 7704.00 frames. ], tot_loss[loss=0.2239, simple_loss=0.3015, pruned_loss=0.07315, over 1610262.46 frames. ], batch size: 18, lr: 5.37e-03, grad_scale: 8.0 +2023-02-06 15:54:11,085 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=112751.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:54:18,473 INFO [optim.py:369] (0/4) 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 15:54:32,355 INFO [train.py:901] (0/4) Epoch 14, batch 7700, loss[loss=0.1688, simple_loss=0.2529, pruned_loss=0.04242, over 7426.00 frames. ], tot_loss[loss=0.2226, simple_loss=0.3002, pruned_loss=0.07245, over 1610704.85 frames. ], batch size: 17, lr: 5.37e-03, grad_scale: 8.0 +2023-02-06 15:54:38,972 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 +2023-02-06 15:54:45,556 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112799.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:54:52,887 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-06 15:54:53,109 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9013, 2.4033, 4.3773, 1.4646, 3.1365, 2.5100, 2.2124, 2.8495], + device='cuda:0'), covar=tensor([0.1818, 0.2550, 0.0716, 0.4610, 0.1749, 0.2889, 0.1903, 0.2624], + device='cuda:0'), in_proj_covar=tensor([0.0498, 0.0552, 0.0539, 0.0601, 0.0623, 0.0565, 0.0491, 0.0622], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 15:55:03,111 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112824.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:55:06,910 INFO [train.py:901] (0/4) Epoch 14, batch 7750, loss[loss=0.3028, simple_loss=0.3546, pruned_loss=0.1255, over 6985.00 frames. ], tot_loss[loss=0.2216, simple_loss=0.2998, pruned_loss=0.07174, over 1611504.47 frames. ], batch size: 72, lr: 5.36e-03, grad_scale: 8.0 +2023-02-06 15:55:24,065 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 +2023-02-06 15:55:28,272 INFO [optim.py:369] (0/4) 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 15:55:30,952 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=112866.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:55:41,025 INFO [train.py:901] (0/4) Epoch 14, batch 7800, loss[loss=0.216, simple_loss=0.304, pruned_loss=0.064, over 8351.00 frames. ], tot_loss[loss=0.2216, simple_loss=0.2997, pruned_loss=0.07181, over 1610395.56 frames. ], batch size: 24, lr: 5.36e-03, grad_scale: 8.0 +2023-02-06 15:55:56,248 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2019, 1.8464, 2.5770, 2.1484, 2.5055, 2.1482, 1.8277, 1.2806], + device='cuda:0'), covar=tensor([0.4641, 0.4384, 0.1450, 0.2658, 0.1963, 0.2481, 0.1810, 0.4595], + device='cuda:0'), in_proj_covar=tensor([0.0900, 0.0909, 0.0750, 0.0874, 0.0947, 0.0835, 0.0718, 0.0785], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 15:56:12,164 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=112924.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:56:13,266 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.59 vs. limit=5.0 +2023-02-06 15:56:14,512 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 +2023-02-06 15:56:16,234 INFO [train.py:901] (0/4) Epoch 14, batch 7850, loss[loss=0.2375, simple_loss=0.3159, pruned_loss=0.07951, over 8315.00 frames. ], tot_loss[loss=0.2207, simple_loss=0.2985, pruned_loss=0.07144, over 1605530.11 frames. ], batch size: 25, lr: 5.36e-03, grad_scale: 8.0 +2023-02-06 15:56:22,337 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=112939.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:56:22,411 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112939.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:56:29,754 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4977, 1.9374, 3.5167, 1.2684, 2.5431, 2.0278, 1.6646, 2.4355], + device='cuda:0'), covar=tensor([0.1745, 0.2532, 0.0567, 0.4286, 0.1592, 0.2800, 0.1939, 0.2145], + device='cuda:0'), in_proj_covar=tensor([0.0496, 0.0551, 0.0535, 0.0600, 0.0622, 0.0564, 0.0489, 0.0622], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 15:56:37,778 INFO [optim.py:369] (0/4) 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 15:56:49,061 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=112979.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:56:49,642 INFO [train.py:901] (0/4) Epoch 14, batch 7900, loss[loss=0.2408, simple_loss=0.3218, pruned_loss=0.07992, over 8659.00 frames. ], tot_loss[loss=0.2216, simple_loss=0.2995, pruned_loss=0.07182, over 1605309.24 frames. ], batch size: 39, lr: 5.36e-03, grad_scale: 8.0 +2023-02-06 15:57:05,614 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2576, 3.0764, 2.2908, 2.5012, 2.5427, 2.1942, 2.5522, 2.7722], + device='cuda:0'), covar=tensor([0.1010, 0.0313, 0.0762, 0.0593, 0.0550, 0.0930, 0.0679, 0.0685], + device='cuda:0'), in_proj_covar=tensor([0.0348, 0.0233, 0.0326, 0.0301, 0.0302, 0.0327, 0.0344, 0.0309], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 15:57:07,094 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-06 15:57:22,229 INFO [train.py:901] (0/4) Epoch 14, batch 7950, loss[loss=0.2292, simple_loss=0.3102, pruned_loss=0.07406, over 8098.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.2993, pruned_loss=0.07148, over 1605697.17 frames. ], batch size: 23, lr: 5.36e-03, grad_scale: 8.0 +2023-02-06 15:57:28,301 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113039.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:57:38,023 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113054.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:57:43,242 INFO [optim.py:369] (0/4) 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 15:57:55,250 INFO [train.py:901] (0/4) Epoch 14, batch 8000, loss[loss=0.259, simple_loss=0.3339, pruned_loss=0.09208, over 7111.00 frames. ], tot_loss[loss=0.2202, simple_loss=0.2988, pruned_loss=0.0708, over 1605653.71 frames. ], batch size: 71, lr: 5.36e-03, grad_scale: 8.0 +2023-02-06 15:58:04,668 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113094.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:58:23,450 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=113122.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:58:28,557 INFO [train.py:901] (0/4) Epoch 14, batch 8050, loss[loss=0.2732, simple_loss=0.3308, pruned_loss=0.1078, over 7275.00 frames. ], tot_loss[loss=0.2213, simple_loss=0.2991, pruned_loss=0.07173, over 1593459.35 frames. ], batch size: 72, lr: 5.36e-03, grad_scale: 16.0 +2023-02-06 15:58:40,112 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=113147.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 15:58:49,769 INFO [optim.py:369] (0/4) 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 15:58:51,437 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/epoch-14.pt +2023-02-06 15:59:02,879 WARNING [train.py:1067] (0/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-06 15:59:06,257 INFO [train.py:901] (0/4) Epoch 15, batch 0, loss[loss=0.2388, simple_loss=0.309, pruned_loss=0.0843, over 8606.00 frames. ], tot_loss[loss=0.2388, simple_loss=0.309, pruned_loss=0.0843, over 8606.00 frames. ], batch size: 34, lr: 5.17e-03, grad_scale: 16.0 +2023-02-06 15:59:06,257 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 15:59:17,270 INFO [train.py:935] (0/4) Epoch 15, validation: loss=0.1825, simple_loss=0.283, pruned_loss=0.04098, over 944034.00 frames. +2023-02-06 15:59:17,271 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-06 15:59:32,300 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-06 15:59:51,451 INFO [train.py:901] (0/4) Epoch 15, batch 50, loss[loss=0.1917, simple_loss=0.2616, pruned_loss=0.06088, over 6791.00 frames. ], tot_loss[loss=0.2175, simple_loss=0.2979, pruned_loss=0.06851, over 367479.44 frames. ], batch size: 15, lr: 5.17e-03, grad_scale: 16.0 +2023-02-06 16:00:08,691 WARNING [train.py:1067] (0/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-06 16:00:21,152 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=113252.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:00:27,787 INFO [optim.py:369] (0/4) 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 16:00:28,489 INFO [train.py:901] (0/4) Epoch 15, batch 100, loss[loss=0.2333, simple_loss=0.3127, pruned_loss=0.07701, over 8258.00 frames. ], tot_loss[loss=0.2198, simple_loss=0.2982, pruned_loss=0.07076, over 641169.90 frames. ], batch size: 24, lr: 5.17e-03, grad_scale: 16.0 +2023-02-06 16:00:29,903 WARNING [train.py:1067] (0/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-06 16:00:41,955 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=113283.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:00:50,246 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=113295.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:01:00,330 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=113310.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:01:02,089 INFO [train.py:901] (0/4) Epoch 15, batch 150, loss[loss=0.1797, simple_loss=0.2577, pruned_loss=0.05085, over 7981.00 frames. ], tot_loss[loss=0.2216, simple_loss=0.3002, pruned_loss=0.07156, over 860591.54 frames. ], batch size: 21, lr: 5.17e-03, grad_scale: 16.0 +2023-02-06 16:01:06,870 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=113320.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:01:16,664 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=113335.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:01:28,763 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=113350.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:01:37,328 INFO [optim.py:369] (0/4) 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 16:01:38,023 INFO [train.py:901] (0/4) Epoch 15, batch 200, loss[loss=0.2117, simple_loss=0.2811, pruned_loss=0.07115, over 7919.00 frames. ], tot_loss[loss=0.2212, simple_loss=0.3001, pruned_loss=0.07113, over 1027136.42 frames. ], batch size: 20, lr: 5.17e-03, grad_scale: 16.0 +2023-02-06 16:01:46,232 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=113375.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:02:01,345 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113398.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:02:11,075 INFO [train.py:901] (0/4) Epoch 15, batch 250, loss[loss=0.2336, simple_loss=0.3076, pruned_loss=0.07979, over 8503.00 frames. ], tot_loss[loss=0.2216, simple_loss=0.3006, pruned_loss=0.07128, over 1161638.51 frames. ], batch size: 26, lr: 5.17e-03, grad_scale: 16.0 +2023-02-06 16:02:19,364 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-06 16:02:28,571 WARNING [train.py:1067] (0/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-06 16:02:43,763 INFO [optim.py:369] (0/4) 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 16:02:44,411 INFO [train.py:901] (0/4) Epoch 15, batch 300, loss[loss=0.2539, simple_loss=0.3357, pruned_loss=0.08604, over 8691.00 frames. ], tot_loss[loss=0.2216, simple_loss=0.3002, pruned_loss=0.07146, over 1263746.83 frames. ], batch size: 34, lr: 5.17e-03, grad_scale: 16.0 +2023-02-06 16:03:19,311 INFO [train.py:901] (0/4) Epoch 15, batch 350, loss[loss=0.198, simple_loss=0.2854, pruned_loss=0.05533, over 8289.00 frames. ], tot_loss[loss=0.2226, simple_loss=0.3014, pruned_loss=0.0719, over 1343890.74 frames. ], batch size: 23, lr: 5.17e-03, grad_scale: 16.0 +2023-02-06 16:03:52,034 INFO [optim.py:369] (0/4) 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 16:03:52,728 INFO [train.py:901] (0/4) Epoch 15, batch 400, loss[loss=0.2159, simple_loss=0.2893, pruned_loss=0.07128, over 7936.00 frames. ], tot_loss[loss=0.2213, simple_loss=0.3003, pruned_loss=0.07114, over 1404595.53 frames. ], batch size: 20, lr: 5.16e-03, grad_scale: 16.0 +2023-02-06 16:04:17,457 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=113596.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:04:17,495 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.8160, 5.8766, 5.1824, 2.3759, 5.3209, 5.6661, 5.5201, 5.3324], + device='cuda:0'), covar=tensor([0.0532, 0.0414, 0.0923, 0.4610, 0.0727, 0.0713, 0.1071, 0.0634], + device='cuda:0'), in_proj_covar=tensor([0.0482, 0.0400, 0.0401, 0.0501, 0.0400, 0.0401, 0.0390, 0.0350], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 16:04:28,807 INFO [train.py:901] (0/4) Epoch 15, batch 450, loss[loss=0.1854, simple_loss=0.2452, pruned_loss=0.06276, over 7528.00 frames. ], tot_loss[loss=0.2203, simple_loss=0.2991, pruned_loss=0.07076, over 1451326.36 frames. ], batch size: 18, lr: 5.16e-03, grad_scale: 16.0 +2023-02-06 16:04:43,280 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=113635.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:04:56,080 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=113654.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:05:01,023 INFO [optim.py:369] (0/4) 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 16:05:01,740 INFO [train.py:901] (0/4) Epoch 15, batch 500, loss[loss=0.2294, simple_loss=0.3073, pruned_loss=0.07572, over 8485.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.2988, pruned_loss=0.07118, over 1485871.44 frames. ], batch size: 29, lr: 5.16e-03, grad_scale: 16.0 +2023-02-06 16:05:02,531 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=113664.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:05:11,071 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2434, 1.4163, 4.4678, 2.0000, 2.3642, 5.0293, 5.0488, 4.3686], + device='cuda:0'), covar=tensor([0.1122, 0.1796, 0.0245, 0.1823, 0.1136, 0.0176, 0.0494, 0.0560], + device='cuda:0'), in_proj_covar=tensor([0.0277, 0.0307, 0.0273, 0.0300, 0.0288, 0.0246, 0.0376, 0.0299], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 16:05:12,301 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=113679.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:05:35,402 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113711.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:05:36,636 INFO [train.py:901] (0/4) Epoch 15, batch 550, loss[loss=0.2007, simple_loss=0.2778, pruned_loss=0.06178, over 7703.00 frames. ], tot_loss[loss=0.2189, simple_loss=0.297, pruned_loss=0.07039, over 1511041.53 frames. ], batch size: 18, lr: 5.16e-03, grad_scale: 16.0 +2023-02-06 16:06:09,825 INFO [optim.py:369] (0/4) 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 16:06:10,536 INFO [train.py:901] (0/4) Epoch 15, batch 600, loss[loss=0.2898, simple_loss=0.3474, pruned_loss=0.1161, over 7153.00 frames. ], tot_loss[loss=0.2202, simple_loss=0.2986, pruned_loss=0.07096, over 1536953.51 frames. ], batch size: 72, lr: 5.16e-03, grad_scale: 16.0 +2023-02-06 16:06:24,218 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 15, batch 650, loss[loss=0.2221, simple_loss=0.3132, pruned_loss=0.06556, over 8291.00 frames. ], tot_loss[loss=0.2208, simple_loss=0.2993, pruned_loss=0.07119, over 1555373.55 frames. ], batch size: 23, lr: 5.16e-03, grad_scale: 16.0 +2023-02-06 16:06:58,553 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.5294, 1.8450, 4.2477, 1.9091, 2.3784, 4.7679, 4.8087, 4.1626], + device='cuda:0'), covar=tensor([0.0991, 0.1643, 0.0303, 0.1960, 0.1225, 0.0192, 0.0385, 0.0543], + device='cuda:0'), in_proj_covar=tensor([0.0279, 0.0309, 0.0274, 0.0301, 0.0288, 0.0248, 0.0377, 0.0300], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 16:07:16,605 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.6234, 3.5536, 3.2333, 2.1332, 3.1535, 3.2366, 3.2969, 3.0002], + device='cuda:0'), covar=tensor([0.0819, 0.0679, 0.0955, 0.3648, 0.0903, 0.1164, 0.1220, 0.0965], + device='cuda:0'), in_proj_covar=tensor([0.0475, 0.0397, 0.0397, 0.0498, 0.0396, 0.0395, 0.0383, 0.0346], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 16:07:19,221 INFO [optim.py:369] (0/4) 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 16:07:19,886 INFO [train.py:901] (0/4) Epoch 15, batch 700, loss[loss=0.2338, simple_loss=0.3036, pruned_loss=0.08204, over 8139.00 frames. ], tot_loss[loss=0.2197, simple_loss=0.2982, pruned_loss=0.0706, over 1567225.21 frames. ], batch size: 22, lr: 5.16e-03, grad_scale: 16.0 +2023-02-06 16:07:37,433 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.8259, 3.8145, 3.4055, 1.8396, 3.3420, 3.4687, 3.3873, 3.2274], + device='cuda:0'), covar=tensor([0.0854, 0.0690, 0.1179, 0.4334, 0.0910, 0.1045, 0.1445, 0.1022], + device='cuda:0'), in_proj_covar=tensor([0.0475, 0.0397, 0.0396, 0.0497, 0.0396, 0.0394, 0.0381, 0.0345], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 16:07:53,445 INFO [train.py:901] (0/4) Epoch 15, batch 750, loss[loss=0.1871, simple_loss=0.2729, pruned_loss=0.05059, over 8110.00 frames. ], tot_loss[loss=0.2208, simple_loss=0.2989, pruned_loss=0.0714, over 1579478.93 frames. ], batch size: 23, lr: 5.16e-03, grad_scale: 16.0 +2023-02-06 16:08:11,149 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-06 16:08:20,460 WARNING [train.py:1067] (0/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-06 16:08:29,183 INFO [optim.py:369] (0/4) 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 16:08:29,872 INFO [train.py:901] (0/4) Epoch 15, batch 800, loss[loss=0.2288, simple_loss=0.3066, pruned_loss=0.07547, over 8482.00 frames. ], tot_loss[loss=0.2217, simple_loss=0.2992, pruned_loss=0.07205, over 1589707.00 frames. ], batch size: 49, lr: 5.16e-03, grad_scale: 16.0 +2023-02-06 16:08:32,827 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=113967.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:08:40,756 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=113979.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:08:47,649 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 +2023-02-06 16:08:50,090 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=113992.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:08:55,521 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-114000.pt +2023-02-06 16:09:01,970 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 15, batch 850, loss[loss=0.2156, simple_loss=0.3022, pruned_loss=0.06448, over 8331.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.2986, pruned_loss=0.07129, over 1597877.35 frames. ], batch size: 26, lr: 5.15e-03, grad_scale: 16.0 +2023-02-06 16:09:21,246 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 +2023-02-06 16:09:39,412 INFO [optim.py:369] (0/4) 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 16:09:40,793 INFO [train.py:901] (0/4) Epoch 15, batch 900, loss[loss=0.2102, simple_loss=0.2786, pruned_loss=0.07087, over 7791.00 frames. ], tot_loss[loss=0.2209, simple_loss=0.2988, pruned_loss=0.07146, over 1602816.06 frames. ], batch size: 19, lr: 5.15e-03, grad_scale: 16.0 +2023-02-06 16:10:02,643 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=114094.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:10:15,158 INFO [train.py:901] (0/4) Epoch 15, batch 950, loss[loss=0.2413, simple_loss=0.311, pruned_loss=0.08581, over 6914.00 frames. ], tot_loss[loss=0.2219, simple_loss=0.3003, pruned_loss=0.07181, over 1605951.33 frames. ], batch size: 71, lr: 5.15e-03, grad_scale: 16.0 +2023-02-06 16:10:20,662 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6867, 2.0791, 2.2654, 1.2335, 2.3660, 1.5894, 0.7321, 1.9192], + device='cuda:0'), covar=tensor([0.0518, 0.0278, 0.0217, 0.0496, 0.0253, 0.0701, 0.0743, 0.0253], + device='cuda:0'), in_proj_covar=tensor([0.0417, 0.0355, 0.0303, 0.0408, 0.0339, 0.0495, 0.0367, 0.0379], + device='cuda:0'), out_proj_covar=tensor([1.1572e-04, 9.5734e-05, 8.1551e-05, 1.1079e-04, 9.2072e-05, 1.4424e-04, + 1.0184e-04, 1.0371e-04], device='cuda:0') +2023-02-06 16:10:21,908 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=114123.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:10:36,818 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=114145.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:10:39,439 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-06 16:10:45,125 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.5677, 4.5454, 4.1550, 2.0953, 4.0429, 4.2129, 4.0865, 3.8810], + device='cuda:0'), covar=tensor([0.0602, 0.0477, 0.0863, 0.4222, 0.0795, 0.0861, 0.1124, 0.0757], + device='cuda:0'), in_proj_covar=tensor([0.0477, 0.0398, 0.0400, 0.0496, 0.0397, 0.0397, 0.0382, 0.0347], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 16:10:49,209 INFO [optim.py:369] (0/4) 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 16:10:49,926 INFO [train.py:901] (0/4) Epoch 15, batch 1000, loss[loss=0.2119, simple_loss=0.2813, pruned_loss=0.07129, over 7538.00 frames. ], tot_loss[loss=0.2224, simple_loss=0.3005, pruned_loss=0.07215, over 1605917.40 frames. ], batch size: 18, lr: 5.15e-03, grad_scale: 16.0 +2023-02-06 16:11:14,210 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-06 16:11:25,580 INFO [train.py:901] (0/4) Epoch 15, batch 1050, loss[loss=0.2219, simple_loss=0.3069, pruned_loss=0.06846, over 8239.00 frames. ], tot_loss[loss=0.2228, simple_loss=0.3011, pruned_loss=0.07227, over 1610292.26 frames. ], batch size: 24, lr: 5.15e-03, grad_scale: 16.0 +2023-02-06 16:11:25,602 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-06 16:11:57,601 INFO [optim.py:369] (0/4) 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 16:11:58,319 INFO [train.py:901] (0/4) Epoch 15, batch 1100, loss[loss=0.2201, simple_loss=0.3036, pruned_loss=0.06832, over 8129.00 frames. ], tot_loss[loss=0.2222, simple_loss=0.3002, pruned_loss=0.07211, over 1610943.31 frames. ], batch size: 22, lr: 5.15e-03, grad_scale: 16.0 +2023-02-06 16:12:26,264 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 +2023-02-06 16:12:33,901 INFO [train.py:901] (0/4) Epoch 15, batch 1150, loss[loss=0.2481, simple_loss=0.3196, pruned_loss=0.08835, over 8104.00 frames. ], tot_loss[loss=0.2237, simple_loss=0.3019, pruned_loss=0.07278, over 1614700.62 frames. ], batch size: 23, lr: 5.15e-03, grad_scale: 16.0 +2023-02-06 16:12:37,515 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.1869, 1.0940, 1.2659, 1.1007, 0.9095, 1.3346, 0.0803, 0.9593], + device='cuda:0'), covar=tensor([0.2113, 0.1592, 0.0594, 0.1020, 0.3292, 0.0534, 0.2672, 0.1497], + device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0179, 0.0111, 0.0218, 0.0259, 0.0115, 0.0165, 0.0176], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 16:12:38,620 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-06 16:12:53,947 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 +2023-02-06 16:12:59,528 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=114350.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:13:07,357 INFO [optim.py:369] (0/4) 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 16:13:07,983 INFO [train.py:901] (0/4) Epoch 15, batch 1200, loss[loss=0.2299, simple_loss=0.2855, pruned_loss=0.08717, over 7433.00 frames. ], tot_loss[loss=0.2221, simple_loss=0.3005, pruned_loss=0.07188, over 1614627.35 frames. ], batch size: 17, lr: 5.15e-03, grad_scale: 16.0 +2023-02-06 16:13:16,129 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=114375.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:13:18,738 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=114379.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:13:36,385 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 15, batch 1250, loss[loss=0.2489, simple_loss=0.3227, pruned_loss=0.08755, over 8470.00 frames. ], tot_loss[loss=0.2232, simple_loss=0.3015, pruned_loss=0.07242, over 1615381.44 frames. ], batch size: 29, lr: 5.15e-03, grad_scale: 16.0 +2023-02-06 16:14:16,860 INFO [optim.py:369] (0/4) 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 16:14:17,472 INFO [train.py:901] (0/4) Epoch 15, batch 1300, loss[loss=0.2117, simple_loss=0.2851, pruned_loss=0.06914, over 7293.00 frames. ], tot_loss[loss=0.2223, simple_loss=0.3008, pruned_loss=0.07191, over 1617403.77 frames. ], batch size: 16, lr: 5.14e-03, grad_scale: 16.0 +2023-02-06 16:14:35,228 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=114489.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:14:51,236 INFO [train.py:901] (0/4) Epoch 15, batch 1350, loss[loss=0.2169, simple_loss=0.2963, pruned_loss=0.06877, over 8029.00 frames. ], tot_loss[loss=0.2216, simple_loss=0.3002, pruned_loss=0.07149, over 1614936.20 frames. ], batch size: 22, lr: 5.14e-03, grad_scale: 16.0 +2023-02-06 16:15:26,453 INFO [optim.py:369] (0/4) 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 16:15:27,119 INFO [train.py:901] (0/4) Epoch 15, batch 1400, loss[loss=0.2459, simple_loss=0.3232, pruned_loss=0.08429, over 8354.00 frames. ], tot_loss[loss=0.2201, simple_loss=0.2992, pruned_loss=0.07048, over 1616091.72 frames. ], batch size: 26, lr: 5.14e-03, grad_scale: 16.0 +2023-02-06 16:15:54,552 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=114604.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:15:54,601 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4786, 1.7746, 1.9161, 1.1628, 1.9031, 1.3625, 0.4050, 1.6782], + device='cuda:0'), covar=tensor([0.0371, 0.0253, 0.0171, 0.0346, 0.0300, 0.0650, 0.0607, 0.0184], + device='cuda:0'), in_proj_covar=tensor([0.0417, 0.0356, 0.0303, 0.0406, 0.0341, 0.0496, 0.0369, 0.0380], + device='cuda:0'), out_proj_covar=tensor([1.1544e-04, 9.6006e-05, 8.1496e-05, 1.1016e-04, 9.2748e-05, 1.4453e-04, + 1.0226e-04, 1.0368e-04], device='cuda:0') +2023-02-06 16:16:00,697 INFO [train.py:901] (0/4) Epoch 15, batch 1450, loss[loss=0.1746, simple_loss=0.2571, pruned_loss=0.0461, over 7664.00 frames. ], tot_loss[loss=0.2207, simple_loss=0.2996, pruned_loss=0.0709, over 1615575.25 frames. ], batch size: 19, lr: 5.14e-03, grad_scale: 16.0 +2023-02-06 16:16:08,829 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-06 16:16:36,179 INFO [optim.py:369] (0/4) 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 16:16:36,893 INFO [train.py:901] (0/4) Epoch 15, batch 1500, loss[loss=0.206, simple_loss=0.2927, pruned_loss=0.05963, over 8439.00 frames. ], tot_loss[loss=0.2204, simple_loss=0.2989, pruned_loss=0.07094, over 1610288.97 frames. ], batch size: 27, lr: 5.14e-03, grad_scale: 16.0 +2023-02-06 16:16:52,687 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.6449, 5.8354, 4.9524, 2.5449, 5.0315, 5.4276, 5.3693, 5.1758], + device='cuda:0'), covar=tensor([0.0591, 0.0402, 0.0968, 0.4084, 0.0719, 0.0792, 0.0966, 0.0477], + device='cuda:0'), in_proj_covar=tensor([0.0483, 0.0398, 0.0400, 0.0496, 0.0394, 0.0397, 0.0383, 0.0349], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 16:17:11,515 INFO [train.py:901] (0/4) Epoch 15, batch 1550, loss[loss=0.2283, simple_loss=0.3126, pruned_loss=0.07199, over 8557.00 frames. ], tot_loss[loss=0.22, simple_loss=0.2983, pruned_loss=0.07083, over 1608661.64 frames. ], batch size: 31, lr: 5.14e-03, grad_scale: 16.0 +2023-02-06 16:17:26,172 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=114734.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:17:45,708 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 15, batch 1600, loss[loss=0.2048, simple_loss=0.274, pruned_loss=0.0678, over 7654.00 frames. ], tot_loss[loss=0.2196, simple_loss=0.2981, pruned_loss=0.07054, over 1607577.93 frames. ], batch size: 19, lr: 5.14e-03, grad_scale: 16.0 +2023-02-06 16:17:58,037 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.8063, 1.9884, 1.8887, 1.9811, 1.0492, 1.6769, 2.0941, 1.7038], + device='cuda:0'), covar=tensor([0.0445, 0.1107, 0.1607, 0.1194, 0.0614, 0.1457, 0.0665, 0.0644], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0152, 0.0189, 0.0156, 0.0102, 0.0161, 0.0115, 0.0138], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 16:18:22,446 INFO [train.py:901] (0/4) Epoch 15, batch 1650, loss[loss=0.2091, simple_loss=0.2918, pruned_loss=0.06321, over 8447.00 frames. ], tot_loss[loss=0.2195, simple_loss=0.2974, pruned_loss=0.07079, over 1607035.35 frames. ], batch size: 25, lr: 5.14e-03, grad_scale: 16.0 +2023-02-06 16:18:55,117 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=114860.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:18:56,272 INFO [optim.py:369] (0/4) 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 16:18:56,966 INFO [train.py:901] (0/4) Epoch 15, batch 1700, loss[loss=0.2211, simple_loss=0.3055, pruned_loss=0.06839, over 8607.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.2978, pruned_loss=0.07054, over 1612225.75 frames. ], batch size: 34, lr: 5.14e-03, grad_scale: 16.0 +2023-02-06 16:19:12,828 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=114885.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:19:16,126 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=114889.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:19:25,772 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6701, 2.0242, 2.3018, 1.0139, 2.3314, 1.4248, 0.7225, 1.8411], + device='cuda:0'), covar=tensor([0.0566, 0.0298, 0.0203, 0.0574, 0.0314, 0.0742, 0.0722, 0.0266], + device='cuda:0'), in_proj_covar=tensor([0.0416, 0.0355, 0.0303, 0.0407, 0.0340, 0.0491, 0.0366, 0.0378], + device='cuda:0'), out_proj_covar=tensor([1.1531e-04, 9.5673e-05, 8.1594e-05, 1.1055e-04, 9.2503e-05, 1.4293e-04, + 1.0128e-04, 1.0329e-04], device='cuda:0') +2023-02-06 16:19:32,924 INFO [train.py:901] (0/4) Epoch 15, batch 1750, loss[loss=0.2204, simple_loss=0.3074, pruned_loss=0.06671, over 8025.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.2978, pruned_loss=0.07016, over 1614683.25 frames. ], batch size: 22, lr: 5.13e-03, grad_scale: 16.0 +2023-02-06 16:20:06,962 INFO [optim.py:369] (0/4) 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 16:20:07,582 INFO [train.py:901] (0/4) Epoch 15, batch 1800, loss[loss=0.2552, simple_loss=0.3234, pruned_loss=0.0935, over 7990.00 frames. ], tot_loss[loss=0.2199, simple_loss=0.2984, pruned_loss=0.07069, over 1613572.61 frames. ], batch size: 21, lr: 5.13e-03, grad_scale: 16.0 +2023-02-06 16:20:43,782 INFO [train.py:901] (0/4) Epoch 15, batch 1850, loss[loss=0.2011, simple_loss=0.2796, pruned_loss=0.06124, over 7969.00 frames. ], tot_loss[loss=0.221, simple_loss=0.2999, pruned_loss=0.07103, over 1616977.30 frames. ], batch size: 21, lr: 5.13e-03, grad_scale: 16.0 +2023-02-06 16:21:17,804 INFO [optim.py:369] (0/4) 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 16:21:18,509 INFO [train.py:901] (0/4) Epoch 15, batch 1900, loss[loss=0.1859, simple_loss=0.2671, pruned_loss=0.05236, over 7649.00 frames. ], tot_loss[loss=0.2201, simple_loss=0.2987, pruned_loss=0.07075, over 1611669.16 frames. ], batch size: 19, lr: 5.13e-03, grad_scale: 16.0 +2023-02-06 16:21:28,838 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=115078.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:21:38,129 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 +2023-02-06 16:21:43,682 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 +2023-02-06 16:21:46,750 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115104.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:21:50,088 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-06 16:21:53,561 INFO [train.py:901] (0/4) Epoch 15, batch 1950, loss[loss=0.2374, simple_loss=0.3218, pruned_loss=0.07648, over 8628.00 frames. ], tot_loss[loss=0.2196, simple_loss=0.2979, pruned_loss=0.07066, over 1612136.20 frames. ], batch size: 39, lr: 5.13e-03, grad_scale: 32.0 +2023-02-06 16:22:04,504 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-06 16:22:23,208 WARNING [train.py:1067] (0/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-06 16:22:28,431 INFO [optim.py:369] (0/4) 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 16:22:29,137 INFO [train.py:901] (0/4) Epoch 15, batch 2000, loss[loss=0.2206, simple_loss=0.3042, pruned_loss=0.06848, over 8455.00 frames. ], tot_loss[loss=0.2187, simple_loss=0.2977, pruned_loss=0.06988, over 1618583.65 frames. ], batch size: 29, lr: 5.13e-03, grad_scale: 32.0 +2023-02-06 16:22:33,856 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0867, 1.7100, 3.6490, 1.6491, 2.4840, 3.9587, 4.0790, 3.4322], + device='cuda:0'), covar=tensor([0.0973, 0.1476, 0.0269, 0.1791, 0.0899, 0.0225, 0.0463, 0.0531], + device='cuda:0'), in_proj_covar=tensor([0.0278, 0.0307, 0.0271, 0.0299, 0.0286, 0.0246, 0.0374, 0.0298], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 16:22:49,761 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=115193.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:22:56,688 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6380, 1.4999, 1.6522, 1.4291, 0.8178, 1.4771, 1.6123, 1.3682], + device='cuda:0'), covar=tensor([0.0513, 0.1223, 0.1725, 0.1345, 0.0591, 0.1503, 0.0658, 0.0646], + device='cuda:0'), in_proj_covar=tensor([0.0098, 0.0153, 0.0190, 0.0156, 0.0102, 0.0161, 0.0115, 0.0138], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 16:23:03,505 INFO [train.py:901] (0/4) Epoch 15, batch 2050, loss[loss=0.2184, simple_loss=0.3089, pruned_loss=0.06398, over 8335.00 frames. ], tot_loss[loss=0.219, simple_loss=0.2982, pruned_loss=0.06992, over 1618343.11 frames. ], batch size: 25, lr: 5.13e-03, grad_scale: 16.0 +2023-02-06 16:23:18,037 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=115233.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:23:39,476 INFO [optim.py:369] (0/4) 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 16:23:39,496 INFO [train.py:901] (0/4) Epoch 15, batch 2100, loss[loss=0.263, simple_loss=0.3468, pruned_loss=0.08955, over 8510.00 frames. ], tot_loss[loss=0.2203, simple_loss=0.2997, pruned_loss=0.07049, over 1618213.05 frames. ], batch size: 28, lr: 5.13e-03, grad_scale: 16.0 +2023-02-06 16:24:13,864 INFO [train.py:901] (0/4) Epoch 15, batch 2150, loss[loss=0.2016, simple_loss=0.2913, pruned_loss=0.05595, over 8029.00 frames. ], tot_loss[loss=0.2209, simple_loss=0.3004, pruned_loss=0.07069, over 1619342.12 frames. ], batch size: 22, lr: 5.13e-03, grad_scale: 16.0 +2023-02-06 16:24:37,886 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=115348.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:24:40,824 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.79 vs. limit=5.0 +2023-02-06 16:24:42,048 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2304, 1.9712, 2.7243, 2.2288, 2.6314, 2.2075, 1.8441, 1.3413], + device='cuda:0'), covar=tensor([0.4632, 0.4432, 0.1376, 0.3009, 0.2078, 0.2534, 0.1858, 0.4626], + device='cuda:0'), in_proj_covar=tensor([0.0901, 0.0912, 0.0753, 0.0884, 0.0944, 0.0835, 0.0716, 0.0791], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 16:24:49,119 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 15, batch 2200, loss[loss=0.1992, simple_loss=0.2673, pruned_loss=0.06553, over 7692.00 frames. ], tot_loss[loss=0.2198, simple_loss=0.2989, pruned_loss=0.07036, over 1611626.50 frames. ], batch size: 18, lr: 5.12e-03, grad_scale: 16.0 +2023-02-06 16:24:59,764 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 +2023-02-06 16:25:07,035 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115388.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:25:07,716 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115389.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:25:24,252 INFO [train.py:901] (0/4) Epoch 15, batch 2250, loss[loss=0.2074, simple_loss=0.2987, pruned_loss=0.05805, over 8040.00 frames. ], tot_loss[loss=0.2202, simple_loss=0.2993, pruned_loss=0.07059, over 1611301.05 frames. ], batch size: 22, lr: 5.12e-03, grad_scale: 8.0 +2023-02-06 16:25:48,112 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=115448.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:25:48,962 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115449.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:25:56,432 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5852, 1.9280, 2.0510, 1.2172, 2.1501, 1.4975, 0.5443, 1.8351], + device='cuda:0'), covar=tensor([0.0459, 0.0255, 0.0172, 0.0415, 0.0264, 0.0657, 0.0678, 0.0214], + device='cuda:0'), in_proj_covar=tensor([0.0420, 0.0359, 0.0307, 0.0414, 0.0343, 0.0499, 0.0373, 0.0385], + device='cuda:0'), out_proj_covar=tensor([1.1645e-04, 9.6753e-05, 8.2531e-05, 1.1242e-04, 9.3360e-05, 1.4535e-04, + 1.0333e-04, 1.0511e-04], device='cuda:0') +2023-02-06 16:25:58,291 INFO [train.py:901] (0/4) Epoch 15, batch 2300, loss[loss=0.2189, simple_loss=0.3098, pruned_loss=0.06402, over 8593.00 frames. ], tot_loss[loss=0.2193, simple_loss=0.2986, pruned_loss=0.06998, over 1614888.24 frames. ], batch size: 49, lr: 5.12e-03, grad_scale: 8.0 +2023-02-06 16:25:58,962 INFO [optim.py:369] (0/4) 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 16:26:07,569 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115474.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:26:22,158 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.0747, 2.7114, 3.7603, 2.0533, 1.9117, 3.7910, 0.8586, 2.1415], + device='cuda:0'), covar=tensor([0.1773, 0.1144, 0.0241, 0.2038, 0.3136, 0.0291, 0.2555, 0.1494], + device='cuda:0'), in_proj_covar=tensor([0.0171, 0.0177, 0.0110, 0.0215, 0.0259, 0.0114, 0.0163, 0.0174], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 16:26:34,675 INFO [train.py:901] (0/4) Epoch 15, batch 2350, loss[loss=0.2166, simple_loss=0.3022, pruned_loss=0.06548, over 8076.00 frames. ], tot_loss[loss=0.2189, simple_loss=0.2976, pruned_loss=0.07006, over 1613931.46 frames. ], batch size: 21, lr: 5.12e-03, grad_scale: 8.0 +2023-02-06 16:26:37,569 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.9864, 1.6412, 1.7812, 1.4950, 1.2417, 1.6054, 2.2679, 1.8146], + device='cuda:0'), covar=tensor([0.0422, 0.1276, 0.1731, 0.1495, 0.0590, 0.1590, 0.0634, 0.0642], + device='cuda:0'), in_proj_covar=tensor([0.0098, 0.0153, 0.0190, 0.0157, 0.0102, 0.0162, 0.0115, 0.0138], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 16:27:02,843 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4294, 2.1417, 2.9251, 2.3885, 2.9090, 2.3125, 1.9521, 1.6035], + device='cuda:0'), covar=tensor([0.4689, 0.4639, 0.1481, 0.3009, 0.1926, 0.2533, 0.1770, 0.4481], + device='cuda:0'), in_proj_covar=tensor([0.0906, 0.0916, 0.0754, 0.0889, 0.0949, 0.0838, 0.0721, 0.0793], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 16:27:09,318 INFO [train.py:901] (0/4) Epoch 15, batch 2400, loss[loss=0.2216, simple_loss=0.2963, pruned_loss=0.07341, over 7817.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.2979, pruned_loss=0.07044, over 1610745.96 frames. ], batch size: 20, lr: 5.12e-03, grad_scale: 8.0 +2023-02-06 16:27:09,502 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=115563.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:27:10,001 INFO [optim.py:369] (0/4) 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 16:27:39,716 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 15, batch 2450, loss[loss=0.2808, simple_loss=0.3284, pruned_loss=0.1166, over 7430.00 frames. ], tot_loss[loss=0.2212, simple_loss=0.2994, pruned_loss=0.07151, over 1616480.00 frames. ], batch size: 71, lr: 5.12e-03, grad_scale: 8.0 +2023-02-06 16:27:53,146 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2070, 2.0977, 1.4494, 1.8616, 1.7806, 1.1158, 1.6724, 1.7454], + device='cuda:0'), covar=tensor([0.1355, 0.0418, 0.1350, 0.0572, 0.0695, 0.1712, 0.0896, 0.0750], + device='cuda:0'), in_proj_covar=tensor([0.0351, 0.0230, 0.0327, 0.0303, 0.0302, 0.0332, 0.0345, 0.0310], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 16:27:56,545 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115629.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:28:19,907 INFO [train.py:901] (0/4) Epoch 15, batch 2500, loss[loss=0.2762, simple_loss=0.3468, pruned_loss=0.1028, over 8582.00 frames. ], tot_loss[loss=0.2205, simple_loss=0.2984, pruned_loss=0.07132, over 1609679.64 frames. ], batch size: 34, lr: 5.12e-03, grad_scale: 8.0 +2023-02-06 16:28:20,558 INFO [optim.py:369] (0/4) 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 16:28:52,057 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3191, 1.7495, 4.6566, 2.0816, 2.4850, 5.2091, 5.2099, 4.4698], + device='cuda:0'), covar=tensor([0.1148, 0.1694, 0.0183, 0.1764, 0.1106, 0.0138, 0.0346, 0.0548], + device='cuda:0'), in_proj_covar=tensor([0.0276, 0.0306, 0.0270, 0.0297, 0.0286, 0.0247, 0.0377, 0.0299], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 16:28:55,237 INFO [train.py:901] (0/4) Epoch 15, batch 2550, loss[loss=0.1834, simple_loss=0.2679, pruned_loss=0.0495, over 8036.00 frames. ], tot_loss[loss=0.2203, simple_loss=0.2985, pruned_loss=0.07098, over 1607832.26 frames. ], batch size: 22, lr: 5.12e-03, grad_scale: 8.0 +2023-02-06 16:29:08,933 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=115732.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:29:09,604 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=115733.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:29:18,332 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2452, 1.9252, 2.6536, 2.1220, 2.4914, 2.1811, 1.8851, 1.2061], + device='cuda:0'), covar=tensor([0.4477, 0.4012, 0.1414, 0.3075, 0.2041, 0.2536, 0.1743, 0.4621], + device='cuda:0'), in_proj_covar=tensor([0.0905, 0.0915, 0.0755, 0.0890, 0.0946, 0.0841, 0.0720, 0.0795], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 16:29:30,413 INFO [train.py:901] (0/4) Epoch 15, batch 2600, loss[loss=0.2053, simple_loss=0.2878, pruned_loss=0.0614, over 7816.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.2978, pruned_loss=0.07024, over 1609615.12 frames. ], batch size: 20, lr: 5.12e-03, grad_scale: 8.0 +2023-02-06 16:29:31,075 INFO [optim.py:369] (0/4) 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 16:29:51,262 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6785, 2.2439, 4.2533, 1.4921, 2.9667, 2.3096, 1.6945, 3.0435], + device='cuda:0'), covar=tensor([0.1763, 0.2411, 0.0678, 0.4210, 0.1636, 0.2827, 0.2179, 0.2207], + device='cuda:0'), in_proj_covar=tensor([0.0497, 0.0551, 0.0541, 0.0602, 0.0622, 0.0566, 0.0497, 0.0618], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 16:30:00,217 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.7257, 2.3381, 3.4787, 2.5539, 3.1273, 2.5247, 2.1370, 1.9760], + device='cuda:0'), covar=tensor([0.4301, 0.4651, 0.1400, 0.3138, 0.2108, 0.2424, 0.1696, 0.4740], + device='cuda:0'), in_proj_covar=tensor([0.0904, 0.0915, 0.0754, 0.0889, 0.0947, 0.0841, 0.0720, 0.0794], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 16:30:04,164 INFO [train.py:901] (0/4) Epoch 15, batch 2650, loss[loss=0.187, simple_loss=0.2702, pruned_loss=0.05188, over 8113.00 frames. ], tot_loss[loss=0.2212, simple_loss=0.2995, pruned_loss=0.07145, over 1613397.09 frames. ], batch size: 23, lr: 5.11e-03, grad_scale: 8.0 +2023-02-06 16:30:07,078 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.1893, 1.4435, 3.3749, 1.1154, 2.9367, 2.8066, 3.0265, 2.9036], + device='cuda:0'), covar=tensor([0.0838, 0.3553, 0.0803, 0.3709, 0.1500, 0.1106, 0.0742, 0.0930], + device='cuda:0'), in_proj_covar=tensor([0.0554, 0.0611, 0.0638, 0.0582, 0.0653, 0.0561, 0.0555, 0.0619], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 16:30:08,489 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115844.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:30:29,375 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=115847.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:30:30,052 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=115848.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:30:32,824 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 +2023-02-06 16:30:39,732 INFO [train.py:901] (0/4) Epoch 15, batch 2700, loss[loss=0.233, simple_loss=0.3116, pruned_loss=0.07724, over 7193.00 frames. ], tot_loss[loss=0.2213, simple_loss=0.2996, pruned_loss=0.07145, over 1614458.37 frames. ], batch size: 72, lr: 5.11e-03, grad_scale: 8.0 +2023-02-06 16:30:40,390 INFO [optim.py:369] (0/4) 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 16:31:12,006 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115910.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 16:31:13,862 INFO [train.py:901] (0/4) Epoch 15, batch 2750, loss[loss=0.1912, simple_loss=0.2727, pruned_loss=0.05491, over 7800.00 frames. ], tot_loss[loss=0.2212, simple_loss=0.2995, pruned_loss=0.07143, over 1616218.05 frames. ], batch size: 20, lr: 5.11e-03, grad_scale: 8.0 +2023-02-06 16:31:49,493 INFO [train.py:901] (0/4) Epoch 15, batch 2800, loss[loss=0.2091, simple_loss=0.2663, pruned_loss=0.07598, over 7542.00 frames. ], tot_loss[loss=0.2214, simple_loss=0.2998, pruned_loss=0.07153, over 1614204.14 frames. ], batch size: 18, lr: 5.11e-03, grad_scale: 8.0 +2023-02-06 16:31:50,140 INFO [optim.py:369] (0/4) 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 16:32:15,112 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-116000.pt +2023-02-06 16:32:18,284 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3118, 2.1653, 1.6375, 1.9099, 1.7910, 1.3859, 1.6686, 1.7058], + device='cuda:0'), covar=tensor([0.1281, 0.0407, 0.1188, 0.0543, 0.0736, 0.1477, 0.0936, 0.0744], + device='cuda:0'), in_proj_covar=tensor([0.0346, 0.0228, 0.0323, 0.0300, 0.0298, 0.0326, 0.0341, 0.0307], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 16:32:24,947 INFO [train.py:901] (0/4) Epoch 15, batch 2850, loss[loss=0.1977, simple_loss=0.2723, pruned_loss=0.06153, over 7786.00 frames. ], tot_loss[loss=0.221, simple_loss=0.2998, pruned_loss=0.07113, over 1615744.99 frames. ], batch size: 19, lr: 5.11e-03, grad_scale: 8.0 +2023-02-06 16:32:38,102 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116032.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:32:42,285 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9395, 1.7060, 2.6564, 1.5623, 2.1863, 2.8814, 2.8601, 2.5905], + device='cuda:0'), covar=tensor([0.0826, 0.1218, 0.0645, 0.1635, 0.1562, 0.0260, 0.0698, 0.0486], + device='cuda:0'), in_proj_covar=tensor([0.0276, 0.0302, 0.0268, 0.0295, 0.0285, 0.0246, 0.0373, 0.0296], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 16:33:00,841 INFO [train.py:901] (0/4) Epoch 15, batch 2900, loss[loss=0.2091, simple_loss=0.3057, pruned_loss=0.05628, over 8104.00 frames. ], tot_loss[loss=0.2215, simple_loss=0.3003, pruned_loss=0.07131, over 1612746.85 frames. ], batch size: 23, lr: 5.11e-03, grad_scale: 8.0 +2023-02-06 16:33:01,415 INFO [optim.py:369] (0/4) 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 16:33:29,127 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116103.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:33:29,811 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116104.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:33:35,327 INFO [train.py:901] (0/4) Epoch 15, batch 2950, loss[loss=0.2057, simple_loss=0.2789, pruned_loss=0.06623, over 7655.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.2997, pruned_loss=0.07127, over 1611824.52 frames. ], batch size: 19, lr: 5.11e-03, grad_scale: 8.0 +2023-02-06 16:33:36,703 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-06 16:33:42,118 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116123.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:33:44,201 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9137, 1.5291, 1.6863, 1.3356, 1.0320, 1.4997, 1.7816, 1.5755], + device='cuda:0'), covar=tensor([0.0524, 0.1184, 0.1597, 0.1381, 0.0606, 0.1428, 0.0664, 0.0606], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0153, 0.0191, 0.0156, 0.0102, 0.0162, 0.0115, 0.0139], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 16:33:45,589 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116128.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:33:46,276 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116134.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:34:08,982 INFO [train.py:901] (0/4) Epoch 15, batch 3000, loss[loss=0.2436, simple_loss=0.3272, pruned_loss=0.07997, over 8321.00 frames. ], tot_loss[loss=0.2219, simple_loss=0.3006, pruned_loss=0.07161, over 1613811.54 frames. ], batch size: 25, lr: 5.11e-03, grad_scale: 8.0 +2023-02-06 16:34:08,982 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 16:34:21,679 INFO [train.py:935] (0/4) 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] (0/4) Maximum memory allocated so far is 6722MB +2023-02-06 16:34:22,354 INFO [optim.py:369] (0/4) 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 16:34:57,896 INFO [train.py:901] (0/4) Epoch 15, batch 3050, loss[loss=0.2187, simple_loss=0.3075, pruned_loss=0.06491, over 8026.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.2994, pruned_loss=0.07087, over 1616752.71 frames. ], batch size: 22, lr: 5.11e-03, grad_scale: 8.0 +2023-02-06 16:35:19,947 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 +2023-02-06 16:35:26,165 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=116254.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 16:35:31,947 INFO [train.py:901] (0/4) Epoch 15, batch 3100, loss[loss=0.2444, simple_loss=0.313, pruned_loss=0.08789, over 8441.00 frames. ], tot_loss[loss=0.2213, simple_loss=0.2995, pruned_loss=0.0715, over 1611928.58 frames. ], batch size: 27, lr: 5.10e-03, grad_scale: 8.0 +2023-02-06 16:35:32,570 INFO [optim.py:369] (0/4) 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 16:35:33,410 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2629, 1.2995, 1.5147, 1.2097, 0.7165, 1.2771, 1.1845, 0.9772], + device='cuda:0'), covar=tensor([0.0544, 0.1227, 0.1576, 0.1325, 0.0562, 0.1464, 0.0679, 0.0667], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0153, 0.0191, 0.0156, 0.0101, 0.0162, 0.0115, 0.0139], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 16:36:06,934 INFO [train.py:901] (0/4) Epoch 15, batch 3150, loss[loss=0.2315, simple_loss=0.3194, pruned_loss=0.07173, over 8664.00 frames. ], tot_loss[loss=0.222, simple_loss=0.3004, pruned_loss=0.07174, over 1615416.53 frames. ], batch size: 34, lr: 5.10e-03, grad_scale: 8.0 +2023-02-06 16:36:20,545 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0698, 1.2693, 1.2085, 0.5390, 1.2350, 0.9958, 0.1358, 1.2036], + device='cuda:0'), covar=tensor([0.0339, 0.0276, 0.0232, 0.0458, 0.0333, 0.0777, 0.0625, 0.0229], + device='cuda:0'), in_proj_covar=tensor([0.0425, 0.0364, 0.0313, 0.0420, 0.0351, 0.0507, 0.0376, 0.0387], + device='cuda:0'), out_proj_covar=tensor([1.1755e-04, 9.7834e-05, 8.3975e-05, 1.1396e-04, 9.5599e-05, 1.4763e-04, + 1.0402e-04, 1.0548e-04], device='cuda:0') +2023-02-06 16:36:27,181 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116341.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:36:30,924 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 +2023-02-06 16:36:34,415 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 +2023-02-06 16:36:41,981 INFO [train.py:901] (0/4) Epoch 15, batch 3200, loss[loss=0.2132, simple_loss=0.2975, pruned_loss=0.06444, over 7426.00 frames. ], tot_loss[loss=0.2236, simple_loss=0.3017, pruned_loss=0.07276, over 1613714.39 frames. ], batch size: 17, lr: 5.10e-03, grad_scale: 8.0 +2023-02-06 16:36:43,354 INFO [optim.py:369] (0/4) 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] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116369.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 16:36:46,960 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 +2023-02-06 16:36:51,221 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=116376.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:37:16,512 INFO [train.py:901] (0/4) Epoch 15, batch 3250, loss[loss=0.2297, simple_loss=0.3032, pruned_loss=0.07808, over 8327.00 frames. ], tot_loss[loss=0.2225, simple_loss=0.3006, pruned_loss=0.07223, over 1610502.71 frames. ], batch size: 26, lr: 5.10e-03, grad_scale: 8.0 +2023-02-06 16:37:52,486 INFO [train.py:901] (0/4) Epoch 15, batch 3300, loss[loss=0.2373, simple_loss=0.3176, pruned_loss=0.07851, over 8583.00 frames. ], tot_loss[loss=0.2214, simple_loss=0.2994, pruned_loss=0.07177, over 1612193.96 frames. ], batch size: 31, lr: 5.10e-03, grad_scale: 8.0 +2023-02-06 16:37:53,156 INFO [optim.py:369] (0/4) 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 16:37:53,278 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116464.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:37:55,220 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=116467.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:38:00,666 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2394, 1.5229, 2.3655, 1.2385, 2.2099, 2.5220, 2.6455, 2.1350], + device='cuda:0'), covar=tensor([0.1158, 0.1179, 0.0444, 0.1932, 0.0663, 0.0399, 0.0708, 0.0721], + device='cuda:0'), in_proj_covar=tensor([0.0277, 0.0305, 0.0270, 0.0299, 0.0287, 0.0248, 0.0376, 0.0299], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 16:38:02,542 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=116478.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:38:12,013 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116491.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:38:26,452 INFO [train.py:901] (0/4) Epoch 15, batch 3350, loss[loss=0.242, simple_loss=0.3413, pruned_loss=0.0714, over 8599.00 frames. ], tot_loss[loss=0.2224, simple_loss=0.3007, pruned_loss=0.07209, over 1617721.05 frames. ], batch size: 39, lr: 5.10e-03, grad_scale: 8.0 +2023-02-06 16:38:33,237 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116523.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:38:45,607 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3708, 1.4959, 1.3316, 1.7705, 0.7031, 1.2482, 1.2872, 1.4498], + device='cuda:0'), covar=tensor([0.0876, 0.0722, 0.1068, 0.0523, 0.1088, 0.1358, 0.0790, 0.0747], + device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0200, 0.0247, 0.0210, 0.0209, 0.0246, 0.0253, 0.0212], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 16:39:02,058 INFO [train.py:901] (0/4) Epoch 15, batch 3400, loss[loss=0.2931, simple_loss=0.3546, pruned_loss=0.1159, over 7180.00 frames. ], tot_loss[loss=0.2221, simple_loss=0.3001, pruned_loss=0.07206, over 1619573.04 frames. ], batch size: 71, lr: 5.10e-03, grad_scale: 8.0 +2023-02-06 16:39:02,719 INFO [optim.py:369] (0/4) 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 16:39:06,250 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.0272, 2.4608, 3.5948, 2.0552, 1.7486, 3.6319, 0.7470, 2.0618], + device='cuda:0'), covar=tensor([0.1574, 0.1338, 0.0220, 0.2126, 0.3676, 0.0279, 0.2825, 0.1787], + device='cuda:0'), in_proj_covar=tensor([0.0171, 0.0178, 0.0110, 0.0214, 0.0258, 0.0114, 0.0162, 0.0176], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 16:39:14,871 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116582.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:39:15,885 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 +2023-02-06 16:39:22,184 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116593.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:39:36,208 INFO [train.py:901] (0/4) Epoch 15, batch 3450, loss[loss=0.2574, simple_loss=0.3274, pruned_loss=0.09364, over 8573.00 frames. ], tot_loss[loss=0.2208, simple_loss=0.2991, pruned_loss=0.07123, over 1619883.29 frames. ], batch size: 49, lr: 5.10e-03, grad_scale: 8.0 +2023-02-06 16:39:44,410 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116625.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 16:39:51,707 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116636.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:40:01,115 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116650.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 16:40:10,183 INFO [train.py:901] (0/4) Epoch 15, batch 3500, loss[loss=0.2438, simple_loss=0.3256, pruned_loss=0.081, over 8432.00 frames. ], tot_loss[loss=0.2218, simple_loss=0.2998, pruned_loss=0.07188, over 1617661.86 frames. ], batch size: 27, lr: 5.10e-03, grad_scale: 8.0 +2023-02-06 16:40:10,857 INFO [optim.py:369] (0/4) 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 16:40:22,993 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 +2023-02-06 16:40:26,403 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=116685.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:40:35,613 WARNING [train.py:1067] (0/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-06 16:40:44,884 INFO [train.py:901] (0/4) Epoch 15, batch 3550, loss[loss=0.1867, simple_loss=0.2796, pruned_loss=0.04695, over 7797.00 frames. ], tot_loss[loss=0.2215, simple_loss=0.2993, pruned_loss=0.0719, over 1612493.06 frames. ], batch size: 19, lr: 5.09e-03, grad_scale: 8.0 +2023-02-06 16:40:46,337 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6632, 1.4458, 1.5471, 1.2426, 0.8681, 1.2932, 1.4568, 1.3224], + device='cuda:0'), covar=tensor([0.0526, 0.1270, 0.1719, 0.1447, 0.0588, 0.1566, 0.0724, 0.0637], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0152, 0.0190, 0.0156, 0.0101, 0.0162, 0.0114, 0.0139], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 16:40:50,941 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116722.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:41:08,597 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116747.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:41:19,416 INFO [train.py:901] (0/4) Epoch 15, batch 3600, loss[loss=0.2597, simple_loss=0.3419, pruned_loss=0.08879, over 8717.00 frames. ], tot_loss[loss=0.2217, simple_loss=0.2998, pruned_loss=0.07176, over 1617860.32 frames. ], batch size: 30, lr: 5.09e-03, grad_scale: 8.0 +2023-02-06 16:41:20,117 INFO [optim.py:369] (0/4) 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 16:41:25,808 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116772.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:41:47,436 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116800.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:41:52,898 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=116808.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:41:56,185 INFO [train.py:901] (0/4) Epoch 15, batch 3650, loss[loss=0.254, simple_loss=0.3455, pruned_loss=0.08127, over 8617.00 frames. ], tot_loss[loss=0.2205, simple_loss=0.2992, pruned_loss=0.0709, over 1614878.24 frames. ], batch size: 49, lr: 5.09e-03, grad_scale: 8.0 +2023-02-06 16:42:00,889 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116820.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:42:08,523 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 +2023-02-06 16:42:13,571 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.6202, 2.8953, 2.5225, 4.0075, 1.6136, 2.1117, 2.3664, 3.1073], + device='cuda:0'), covar=tensor([0.0639, 0.0830, 0.0816, 0.0253, 0.1175, 0.1297, 0.1062, 0.0773], + device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0203, 0.0251, 0.0212, 0.0211, 0.0249, 0.0255, 0.0213], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 16:42:13,596 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116838.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:42:21,036 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116849.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:42:24,469 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5001, 2.5923, 1.7991, 2.1869, 2.1046, 1.4942, 1.8547, 2.0793], + device='cuda:0'), covar=tensor([0.1392, 0.0340, 0.1101, 0.0620, 0.0695, 0.1501, 0.0987, 0.0828], + device='cuda:0'), in_proj_covar=tensor([0.0353, 0.0232, 0.0327, 0.0305, 0.0302, 0.0333, 0.0348, 0.0310], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 16:42:30,281 INFO [train.py:901] (0/4) Epoch 15, batch 3700, loss[loss=0.2621, simple_loss=0.3411, pruned_loss=0.09159, over 8343.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.301, pruned_loss=0.07224, over 1612894.43 frames. ], batch size: 25, lr: 5.09e-03, grad_scale: 8.0 +2023-02-06 16:42:30,476 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 16:42:33,136 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=116867.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:42:36,571 WARNING [train.py:1067] (0/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-06 16:42:38,067 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116874.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:43:06,653 INFO [train.py:901] (0/4) Epoch 15, batch 3750, loss[loss=0.1891, simple_loss=0.2748, pruned_loss=0.05166, over 7807.00 frames. ], tot_loss[loss=0.221, simple_loss=0.2996, pruned_loss=0.07118, over 1613948.00 frames. ], batch size: 20, lr: 5.09e-03, grad_scale: 8.0 +2023-02-06 16:43:13,675 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116923.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:43:14,654 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-02-06 16:43:40,817 INFO [train.py:901] (0/4) Epoch 15, batch 3800, loss[loss=0.2665, simple_loss=0.3474, pruned_loss=0.09274, over 8509.00 frames. ], tot_loss[loss=0.2205, simple_loss=0.2991, pruned_loss=0.07093, over 1612185.01 frames. ], batch size: 26, lr: 5.09e-03, grad_scale: 8.0 +2023-02-06 16:43:41,467 INFO [optim.py:369] (0/4) 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 16:43:52,408 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=116980.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:43:53,903 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116982.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:44:15,603 INFO [train.py:901] (0/4) Epoch 15, batch 3850, loss[loss=0.2203, simple_loss=0.3054, pruned_loss=0.0676, over 8470.00 frames. ], tot_loss[loss=0.2186, simple_loss=0.297, pruned_loss=0.07016, over 1609026.79 frames. ], batch size: 29, lr: 5.09e-03, grad_scale: 8.0 +2023-02-06 16:44:26,332 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 +2023-02-06 16:44:42,571 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-06 16:44:46,230 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=117056.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:44:50,982 INFO [train.py:901] (0/4) Epoch 15, batch 3900, loss[loss=0.241, simple_loss=0.321, pruned_loss=0.08054, over 8439.00 frames. ], tot_loss[loss=0.2182, simple_loss=0.2966, pruned_loss=0.06991, over 1608228.37 frames. ], batch size: 27, lr: 5.09e-03, grad_scale: 8.0 +2023-02-06 16:44:51,616 INFO [optim.py:369] (0/4) 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 16:44:53,026 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=117066.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:45:03,680 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=117081.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:45:12,944 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117095.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:45:24,919 INFO [train.py:901] (0/4) Epoch 15, batch 3950, loss[loss=0.2131, simple_loss=0.2932, pruned_loss=0.06656, over 8364.00 frames. ], tot_loss[loss=0.2199, simple_loss=0.2978, pruned_loss=0.07094, over 1610617.85 frames. ], batch size: 24, lr: 5.09e-03, grad_scale: 8.0 +2023-02-06 16:46:01,088 INFO [train.py:901] (0/4) Epoch 15, batch 4000, loss[loss=0.211, simple_loss=0.2912, pruned_loss=0.06539, over 7649.00 frames. ], tot_loss[loss=0.2192, simple_loss=0.2972, pruned_loss=0.07055, over 1611954.78 frames. ], batch size: 19, lr: 5.08e-03, grad_scale: 8.0 +2023-02-06 16:46:01,785 INFO [optim.py:369] (0/4) 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 16:46:01,867 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=117164.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:46:12,553 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=117179.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:46:13,875 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117181.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:46:29,831 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 15, batch 4050, loss[loss=0.2376, simple_loss=0.3104, pruned_loss=0.08241, over 8238.00 frames. ], tot_loss[loss=0.221, simple_loss=0.2988, pruned_loss=0.07165, over 1611851.75 frames. ], batch size: 24, lr: 5.08e-03, grad_scale: 8.0 +2023-02-06 16:46:47,321 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-06 16:46:47,801 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8845, 1.7172, 3.5002, 1.5524, 2.3137, 3.8386, 3.9594, 3.2249], + device='cuda:0'), covar=tensor([0.1236, 0.1643, 0.0332, 0.2114, 0.1110, 0.0251, 0.0532, 0.0619], + device='cuda:0'), in_proj_covar=tensor([0.0278, 0.0306, 0.0270, 0.0302, 0.0287, 0.0248, 0.0377, 0.0299], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 16:46:53,286 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=117238.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:46:53,873 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=117239.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:47:11,628 INFO [train.py:901] (0/4) Epoch 15, batch 4100, loss[loss=0.2087, simple_loss=0.2759, pruned_loss=0.0707, over 7911.00 frames. ], tot_loss[loss=0.2216, simple_loss=0.2995, pruned_loss=0.07186, over 1612016.16 frames. ], batch size: 20, lr: 5.08e-03, grad_scale: 8.0 +2023-02-06 16:47:11,833 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=117263.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:47:12,284 INFO [optim.py:369] (0/4) 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 16:47:22,912 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 15, batch 4150, loss[loss=0.1975, simple_loss=0.2779, pruned_loss=0.05853, over 7947.00 frames. ], tot_loss[loss=0.2193, simple_loss=0.2973, pruned_loss=0.07071, over 1607597.51 frames. ], batch size: 20, lr: 5.08e-03, grad_scale: 8.0 +2023-02-06 16:48:10,069 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=117347.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:48:12,983 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=117351.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:48:21,503 INFO [train.py:901] (0/4) Epoch 15, batch 4200, loss[loss=0.2068, simple_loss=0.2978, pruned_loss=0.05795, over 8246.00 frames. ], tot_loss[loss=0.2203, simple_loss=0.2987, pruned_loss=0.07098, over 1613810.04 frames. ], batch size: 24, lr: 5.08e-03, grad_scale: 8.0 +2023-02-06 16:48:22,822 INFO [optim.py:369] (0/4) 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 16:48:31,967 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=117376.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:48:40,581 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-06 16:48:57,043 INFO [train.py:901] (0/4) Epoch 15, batch 4250, loss[loss=0.2074, simple_loss=0.2994, pruned_loss=0.05773, over 8466.00 frames. ], tot_loss[loss=0.2201, simple_loss=0.2986, pruned_loss=0.07083, over 1613114.12 frames. ], batch size: 25, lr: 5.08e-03, grad_scale: 16.0 +2023-02-06 16:49:03,728 WARNING [train.py:1067] (0/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-06 16:49:14,096 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=117437.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:49:17,734 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-06 16:49:30,903 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=117462.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:49:31,425 INFO [train.py:901] (0/4) Epoch 15, batch 4300, loss[loss=0.2141, simple_loss=0.3004, pruned_loss=0.06393, over 8292.00 frames. ], tot_loss[loss=0.2198, simple_loss=0.2982, pruned_loss=0.07069, over 1612217.80 frames. ], batch size: 23, lr: 5.08e-03, grad_scale: 16.0 +2023-02-06 16:49:32,091 INFO [optim.py:369] (0/4) 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 16:49:59,158 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 +2023-02-06 16:50:05,915 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 +2023-02-06 16:50:07,601 INFO [train.py:901] (0/4) Epoch 15, batch 4350, loss[loss=0.2231, simple_loss=0.2966, pruned_loss=0.07474, over 7977.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.2993, pruned_loss=0.07098, over 1616030.70 frames. ], batch size: 21, lr: 5.08e-03, grad_scale: 16.0 +2023-02-06 16:50:22,377 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.7337, 2.4211, 3.4735, 2.8159, 3.1926, 2.6696, 2.2157, 2.0030], + device='cuda:0'), covar=tensor([0.4237, 0.4581, 0.1327, 0.2854, 0.1961, 0.2355, 0.1736, 0.4616], + device='cuda:0'), in_proj_covar=tensor([0.0908, 0.0915, 0.0752, 0.0880, 0.0954, 0.0839, 0.0717, 0.0790], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 16:50:23,043 INFO [zipformer.py:1185] (0/4) 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] (0/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-06 16:50:40,548 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=117560.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:50:42,371 INFO [train.py:901] (0/4) Epoch 15, batch 4400, loss[loss=0.2437, simple_loss=0.3235, pruned_loss=0.08195, over 8316.00 frames. ], tot_loss[loss=0.2222, simple_loss=0.3004, pruned_loss=0.07195, over 1619038.74 frames. ], batch size: 25, lr: 5.08e-03, grad_scale: 16.0 +2023-02-06 16:50:43,039 INFO [optim.py:369] (0/4) 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 16:50:55,784 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=117583.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:51:16,413 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 +2023-02-06 16:51:17,967 INFO [train.py:901] (0/4) Epoch 15, batch 4450, loss[loss=0.2546, simple_loss=0.3239, pruned_loss=0.09263, over 8029.00 frames. ], tot_loss[loss=0.2226, simple_loss=0.3005, pruned_loss=0.07233, over 1612276.04 frames. ], batch size: 22, lr: 5.07e-03, grad_scale: 16.0 +2023-02-06 16:51:17,986 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-06 16:51:48,982 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5274, 1.8557, 2.0175, 1.1895, 2.0992, 1.3986, 0.5267, 1.7814], + device='cuda:0'), covar=tensor([0.0506, 0.0281, 0.0220, 0.0443, 0.0280, 0.0742, 0.0705, 0.0240], + device='cuda:0'), in_proj_covar=tensor([0.0415, 0.0358, 0.0309, 0.0412, 0.0343, 0.0503, 0.0372, 0.0378], + device='cuda:0'), out_proj_covar=tensor([1.1465e-04, 9.6427e-05, 8.2773e-05, 1.1158e-04, 9.3190e-05, 1.4649e-04, + 1.0254e-04, 1.0265e-04], device='cuda:0') +2023-02-06 16:51:52,100 INFO [train.py:901] (0/4) Epoch 15, batch 4500, loss[loss=0.1742, simple_loss=0.2594, pruned_loss=0.04446, over 7807.00 frames. ], tot_loss[loss=0.2214, simple_loss=0.2999, pruned_loss=0.07149, over 1616815.86 frames. ], batch size: 19, lr: 5.07e-03, grad_scale: 16.0 +2023-02-06 16:51:52,735 INFO [optim.py:369] (0/4) 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 16:52:11,212 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=117691.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:52:11,853 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-06 16:52:16,197 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117698.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:52:27,013 INFO [train.py:901] (0/4) Epoch 15, batch 4550, loss[loss=0.2113, simple_loss=0.2992, pruned_loss=0.06171, over 8350.00 frames. ], tot_loss[loss=0.2196, simple_loss=0.298, pruned_loss=0.07061, over 1609876.17 frames. ], batch size: 24, lr: 5.07e-03, grad_scale: 8.0 +2023-02-06 16:52:31,115 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.9039, 5.9828, 5.1946, 2.4033, 5.3038, 5.6312, 5.5410, 5.4137], + device='cuda:0'), covar=tensor([0.0535, 0.0401, 0.0881, 0.4902, 0.0691, 0.0859, 0.1181, 0.0688], + device='cuda:0'), in_proj_covar=tensor([0.0483, 0.0401, 0.0403, 0.0501, 0.0394, 0.0403, 0.0386, 0.0348], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 16:53:02,105 INFO [train.py:901] (0/4) Epoch 15, batch 4600, loss[loss=0.1995, simple_loss=0.2833, pruned_loss=0.05788, over 7655.00 frames. ], tot_loss[loss=0.2189, simple_loss=0.2973, pruned_loss=0.07028, over 1610651.31 frames. ], batch size: 19, lr: 5.07e-03, grad_scale: 8.0 +2023-02-06 16:53:03,474 INFO [optim.py:369] (0/4) 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 16:53:31,590 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.7878, 5.8222, 5.2317, 2.5000, 5.1599, 5.5399, 5.3724, 5.3427], + device='cuda:0'), covar=tensor([0.0520, 0.0416, 0.0970, 0.4663, 0.0715, 0.0723, 0.1186, 0.0513], + device='cuda:0'), in_proj_covar=tensor([0.0487, 0.0404, 0.0408, 0.0506, 0.0398, 0.0407, 0.0389, 0.0351], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 16:53:31,670 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117806.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:53:36,043 INFO [train.py:901] (0/4) Epoch 15, batch 4650, loss[loss=0.214, simple_loss=0.2941, pruned_loss=0.06692, over 8497.00 frames. ], tot_loss[loss=0.2189, simple_loss=0.2975, pruned_loss=0.07015, over 1607873.48 frames. ], batch size: 26, lr: 5.07e-03, grad_scale: 8.0 +2023-02-06 16:54:11,643 INFO [train.py:901] (0/4) Epoch 15, batch 4700, loss[loss=0.2233, simple_loss=0.3, pruned_loss=0.07332, over 8092.00 frames. ], tot_loss[loss=0.2184, simple_loss=0.2969, pruned_loss=0.06997, over 1611315.82 frames. ], batch size: 21, lr: 5.07e-03, grad_scale: 8.0 +2023-02-06 16:54:12,892 INFO [optim.py:369] (0/4) 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 16:54:46,542 INFO [train.py:901] (0/4) Epoch 15, batch 4750, loss[loss=0.2207, simple_loss=0.3006, pruned_loss=0.07043, over 8130.00 frames. ], tot_loss[loss=0.2205, simple_loss=0.2987, pruned_loss=0.07117, over 1608474.60 frames. ], batch size: 22, lr: 5.07e-03, grad_scale: 8.0 +2023-02-06 16:55:11,957 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-06 16:55:15,287 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-06 16:55:16,043 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=117954.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:55:22,520 INFO [train.py:901] (0/4) Epoch 15, batch 4800, loss[loss=0.2585, simple_loss=0.3419, pruned_loss=0.08753, over 8633.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.2989, pruned_loss=0.07117, over 1613541.84 frames. ], batch size: 31, lr: 5.07e-03, grad_scale: 8.0 +2023-02-06 16:55:23,941 INFO [optim.py:369] (0/4) 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 16:55:33,833 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=117979.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:55:47,559 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-118000.pt +2023-02-06 16:55:57,738 INFO [train.py:901] (0/4) Epoch 15, batch 4850, loss[loss=0.2157, simple_loss=0.299, pruned_loss=0.06618, over 8026.00 frames. ], tot_loss[loss=0.2219, simple_loss=0.3, pruned_loss=0.07186, over 1615572.38 frames. ], batch size: 22, lr: 5.07e-03, grad_scale: 8.0 +2023-02-06 16:56:07,046 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-06 16:56:29,208 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118058.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:56:31,984 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118062.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:56:32,449 INFO [train.py:901] (0/4) Epoch 15, batch 4900, loss[loss=0.2098, simple_loss=0.2812, pruned_loss=0.06923, over 7683.00 frames. ], tot_loss[loss=0.2229, simple_loss=0.3006, pruned_loss=0.07259, over 1618646.68 frames. ], batch size: 18, lr: 5.07e-03, grad_scale: 8.0 +2023-02-06 16:56:33,728 INFO [optim.py:369] (0/4) 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 16:56:50,265 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118087.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:57:03,285 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.38 vs. limit=5.0 +2023-02-06 16:57:04,828 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.12 vs. limit=5.0 +2023-02-06 16:57:07,608 INFO [train.py:901] (0/4) Epoch 15, batch 4950, loss[loss=0.1959, simple_loss=0.2716, pruned_loss=0.06012, over 8245.00 frames. ], tot_loss[loss=0.2215, simple_loss=0.2996, pruned_loss=0.07169, over 1619934.91 frames. ], batch size: 22, lr: 5.06e-03, grad_scale: 8.0 +2023-02-06 16:57:38,290 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9188, 1.7431, 3.3605, 1.5395, 2.4171, 3.7934, 3.8034, 3.1746], + device='cuda:0'), covar=tensor([0.1197, 0.1577, 0.0363, 0.2070, 0.0957, 0.0230, 0.0498, 0.0620], + device='cuda:0'), in_proj_covar=tensor([0.0274, 0.0304, 0.0268, 0.0299, 0.0287, 0.0246, 0.0373, 0.0298], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 16:57:42,115 INFO [train.py:901] (0/4) Epoch 15, batch 5000, loss[loss=0.2084, simple_loss=0.2963, pruned_loss=0.0602, over 8256.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.2996, pruned_loss=0.07131, over 1619407.57 frames. ], batch size: 24, lr: 5.06e-03, grad_scale: 8.0 +2023-02-06 16:57:43,373 INFO [optim.py:369] (0/4) 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 16:57:58,591 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118186.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:58:17,614 INFO [train.py:901] (0/4) Epoch 15, batch 5050, loss[loss=0.2187, simple_loss=0.2972, pruned_loss=0.07009, over 7931.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.2993, pruned_loss=0.07144, over 1613323.54 frames. ], batch size: 20, lr: 5.06e-03, grad_scale: 8.0 +2023-02-06 16:58:22,822 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 +2023-02-06 16:58:43,432 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-06 16:58:52,555 INFO [train.py:901] (0/4) Epoch 15, batch 5100, loss[loss=0.1907, simple_loss=0.2725, pruned_loss=0.05439, over 7976.00 frames. ], tot_loss[loss=0.222, simple_loss=0.2999, pruned_loss=0.07198, over 1612318.61 frames. ], batch size: 21, lr: 5.06e-03, grad_scale: 8.0 +2023-02-06 16:58:53,829 INFO [optim.py:369] (0/4) 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 16:59:08,396 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118287.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:59:18,594 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-06 16:59:23,821 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118307.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:59:27,363 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2248, 2.0559, 2.8538, 2.2625, 2.6621, 2.2246, 1.9151, 1.5063], + device='cuda:0'), covar=tensor([0.4859, 0.4685, 0.1609, 0.3291, 0.2356, 0.2763, 0.1915, 0.4994], + device='cuda:0'), in_proj_covar=tensor([0.0910, 0.0915, 0.0754, 0.0885, 0.0955, 0.0838, 0.0717, 0.0795], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 16:59:27,775 INFO [train.py:901] (0/4) Epoch 15, batch 5150, loss[loss=0.2512, simple_loss=0.3267, pruned_loss=0.08791, over 8501.00 frames. ], tot_loss[loss=0.2216, simple_loss=0.2997, pruned_loss=0.07177, over 1610412.57 frames. ], batch size: 28, lr: 5.06e-03, grad_scale: 8.0 +2023-02-06 16:59:51,072 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118347.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 16:59:51,389 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 +2023-02-06 17:00:02,420 INFO [train.py:901] (0/4) Epoch 15, batch 5200, loss[loss=0.2025, simple_loss=0.2906, pruned_loss=0.05717, over 8245.00 frames. ], tot_loss[loss=0.2224, simple_loss=0.3003, pruned_loss=0.0723, over 1614489.98 frames. ], batch size: 22, lr: 5.06e-03, grad_scale: 8.0 +2023-02-06 17:00:03,701 INFO [optim.py:369] (0/4) 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 17:00:29,681 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=118402.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:00:37,923 INFO [train.py:901] (0/4) Epoch 15, batch 5250, loss[loss=0.2156, simple_loss=0.2955, pruned_loss=0.06787, over 8480.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.299, pruned_loss=0.07109, over 1613843.78 frames. ], batch size: 26, lr: 5.06e-03, grad_scale: 8.0 +2023-02-06 17:00:46,152 WARNING [train.py:1067] (0/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-06 17:01:12,971 INFO [train.py:901] (0/4) Epoch 15, batch 5300, loss[loss=0.2172, simple_loss=0.2937, pruned_loss=0.0704, over 8619.00 frames. ], tot_loss[loss=0.2205, simple_loss=0.2989, pruned_loss=0.07109, over 1615688.76 frames. ], batch size: 34, lr: 5.06e-03, grad_scale: 8.0 +2023-02-06 17:01:14,347 INFO [optim.py:369] (0/4) 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 17:01:47,925 INFO [train.py:901] (0/4) Epoch 15, batch 5350, loss[loss=0.2222, simple_loss=0.3088, pruned_loss=0.06783, over 8330.00 frames. ], tot_loss[loss=0.2205, simple_loss=0.2988, pruned_loss=0.07109, over 1613421.39 frames. ], batch size: 25, lr: 5.06e-03, grad_scale: 8.0 +2023-02-06 17:01:50,867 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118517.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:02:01,049 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=118530.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:02:15,660 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.9905, 2.3835, 3.5932, 1.8624, 1.6184, 3.5588, 0.7158, 1.9939], + device='cuda:0'), covar=tensor([0.1741, 0.1441, 0.0212, 0.2141, 0.3496, 0.0322, 0.2909, 0.1740], + device='cuda:0'), in_proj_covar=tensor([0.0171, 0.0177, 0.0111, 0.0212, 0.0256, 0.0115, 0.0160, 0.0174], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 17:02:24,465 INFO [train.py:901] (0/4) Epoch 15, batch 5400, loss[loss=0.1888, simple_loss=0.2724, pruned_loss=0.0526, over 8481.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.2995, pruned_loss=0.07137, over 1613588.78 frames. ], batch size: 25, lr: 5.05e-03, grad_scale: 8.0 +2023-02-06 17:02:25,794 INFO [optim.py:369] (0/4) 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 17:02:58,969 INFO [train.py:901] (0/4) Epoch 15, batch 5450, loss[loss=0.2161, simple_loss=0.2932, pruned_loss=0.06944, over 8440.00 frames. ], tot_loss[loss=0.2218, simple_loss=0.3003, pruned_loss=0.07163, over 1617999.93 frames. ], batch size: 49, lr: 5.05e-03, grad_scale: 8.0 +2023-02-06 17:03:11,222 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=118631.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:03:21,575 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118645.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:03:22,326 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3595, 1.4758, 1.4966, 1.1212, 1.5570, 1.2022, 0.6191, 1.4208], + device='cuda:0'), covar=tensor([0.0360, 0.0235, 0.0169, 0.0340, 0.0259, 0.0500, 0.0506, 0.0185], + device='cuda:0'), in_proj_covar=tensor([0.0419, 0.0361, 0.0310, 0.0415, 0.0346, 0.0508, 0.0373, 0.0383], + device='cuda:0'), out_proj_covar=tensor([1.1577e-04, 9.7042e-05, 8.3039e-05, 1.1209e-04, 9.3657e-05, 1.4806e-04, + 1.0268e-04, 1.0398e-04], device='cuda:0') +2023-02-06 17:03:24,936 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118649.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:03:26,158 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=118651.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:03:34,864 INFO [train.py:901] (0/4) Epoch 15, batch 5500, loss[loss=0.1659, simple_loss=0.2438, pruned_loss=0.04403, over 7546.00 frames. ], tot_loss[loss=0.2196, simple_loss=0.2982, pruned_loss=0.07048, over 1611709.44 frames. ], batch size: 18, lr: 5.05e-03, grad_scale: 8.0 +2023-02-06 17:03:36,239 INFO [optim.py:369] (0/4) 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 17:03:38,415 WARNING [train.py:1067] (0/4) 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] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=118691.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:04:09,094 INFO [train.py:901] (0/4) Epoch 15, batch 5550, loss[loss=0.1701, simple_loss=0.2598, pruned_loss=0.04022, over 7650.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.2968, pruned_loss=0.0699, over 1607514.52 frames. ], batch size: 19, lr: 5.05e-03, grad_scale: 8.0 +2023-02-06 17:04:32,278 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118746.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:04:44,981 INFO [train.py:901] (0/4) Epoch 15, batch 5600, loss[loss=0.22, simple_loss=0.2994, pruned_loss=0.07031, over 8318.00 frames. ], tot_loss[loss=0.2195, simple_loss=0.2982, pruned_loss=0.07038, over 1610932.37 frames. ], batch size: 25, lr: 5.05e-03, grad_scale: 8.0 +2023-02-06 17:04:46,299 INFO [optim.py:369] (0/4) 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 17:04:47,197 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118766.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:04:52,580 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118773.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:04:55,257 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.9354, 2.4534, 3.4375, 1.7983, 1.7385, 3.3989, 0.6345, 1.9550], + device='cuda:0'), covar=tensor([0.1665, 0.1290, 0.0304, 0.2249, 0.3344, 0.0297, 0.3190, 0.1790], + device='cuda:0'), in_proj_covar=tensor([0.0171, 0.0177, 0.0111, 0.0211, 0.0255, 0.0115, 0.0160, 0.0173], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 17:05:09,079 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118798.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:05:09,671 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.7926, 5.8535, 5.1769, 2.3841, 5.2183, 5.5607, 5.4700, 5.2475], + device='cuda:0'), covar=tensor([0.0584, 0.0434, 0.0969, 0.4718, 0.0661, 0.0766, 0.1016, 0.0581], + device='cuda:0'), in_proj_covar=tensor([0.0491, 0.0406, 0.0412, 0.0510, 0.0401, 0.0414, 0.0391, 0.0357], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 17:05:14,505 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 15, batch 5650, loss[loss=0.2521, simple_loss=0.3331, pruned_loss=0.08558, over 8517.00 frames. ], tot_loss[loss=0.221, simple_loss=0.2994, pruned_loss=0.07127, over 1616730.53 frames. ], batch size: 28, lr: 5.05e-03, grad_scale: 8.0 +2023-02-06 17:05:43,425 WARNING [train.py:1067] (0/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-06 17:05:53,492 INFO [train.py:901] (0/4) Epoch 15, batch 5700, loss[loss=0.2066, simple_loss=0.2805, pruned_loss=0.06635, over 7540.00 frames. ], tot_loss[loss=0.2213, simple_loss=0.2996, pruned_loss=0.07153, over 1612561.45 frames. ], batch size: 18, lr: 5.05e-03, grad_scale: 8.0 +2023-02-06 17:05:54,817 INFO [optim.py:369] (0/4) 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 17:06:14,682 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 +2023-02-06 17:06:21,156 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118901.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:06:29,110 INFO [train.py:901] (0/4) Epoch 15, batch 5750, loss[loss=0.1931, simple_loss=0.2854, pruned_loss=0.05034, over 8668.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.301, pruned_loss=0.07217, over 1611112.98 frames. ], batch size: 34, lr: 5.05e-03, grad_scale: 8.0 +2023-02-06 17:06:38,206 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118926.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:06:46,380 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-06 17:07:04,205 INFO [train.py:901] (0/4) Epoch 15, batch 5800, loss[loss=0.2257, simple_loss=0.3096, pruned_loss=0.07089, over 8456.00 frames. ], tot_loss[loss=0.2212, simple_loss=0.2999, pruned_loss=0.07131, over 1612267.33 frames. ], batch size: 27, lr: 5.05e-03, grad_scale: 8.0 +2023-02-06 17:07:05,539 INFO [optim.py:369] (0/4) 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] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=118993.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:07:32,094 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119002.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:07:39,884 INFO [train.py:901] (0/4) Epoch 15, batch 5850, loss[loss=0.2018, simple_loss=0.2677, pruned_loss=0.06796, over 7285.00 frames. ], tot_loss[loss=0.2213, simple_loss=0.2994, pruned_loss=0.07154, over 1612024.30 frames. ], batch size: 16, lr: 5.04e-03, grad_scale: 8.0 +2023-02-06 17:07:46,174 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119022.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:07:49,451 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119027.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:08:02,794 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119047.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:08:13,672 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 15, batch 5900, loss[loss=0.1894, simple_loss=0.275, pruned_loss=0.05185, over 8125.00 frames. ], tot_loss[loss=0.2213, simple_loss=0.3001, pruned_loss=0.07128, over 1614921.88 frames. ], batch size: 22, lr: 5.04e-03, grad_scale: 8.0 +2023-02-06 17:08:15,366 INFO [optim.py:369] (0/4) 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 17:08:30,146 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119087.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:08:34,134 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=119108.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:08:45,485 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4936, 1.8200, 1.8923, 0.9435, 1.9392, 1.3767, 0.4182, 1.6554], + device='cuda:0'), covar=tensor([0.0443, 0.0283, 0.0225, 0.0502, 0.0294, 0.0730, 0.0678, 0.0254], + device='cuda:0'), in_proj_covar=tensor([0.0416, 0.0359, 0.0309, 0.0415, 0.0344, 0.0502, 0.0373, 0.0382], + device='cuda:0'), out_proj_covar=tensor([1.1483e-04, 9.6403e-05, 8.2617e-05, 1.1202e-04, 9.3162e-05, 1.4612e-04, + 1.0260e-04, 1.0383e-04], device='cuda:0') +2023-02-06 17:08:46,716 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 15, batch 5950, loss[loss=0.2048, simple_loss=0.2885, pruned_loss=0.06058, over 8250.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.2999, pruned_loss=0.07113, over 1609890.61 frames. ], batch size: 24, lr: 5.04e-03, grad_scale: 8.0 +2023-02-06 17:09:22,897 INFO [train.py:901] (0/4) Epoch 15, batch 6000, loss[loss=0.2312, simple_loss=0.2956, pruned_loss=0.08347, over 7533.00 frames. ], tot_loss[loss=0.2205, simple_loss=0.2986, pruned_loss=0.07115, over 1606334.40 frames. ], batch size: 18, lr: 5.04e-03, grad_scale: 8.0 +2023-02-06 17:09:22,898 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 17:09:35,270 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1214, 1.8245, 2.4887, 2.0388, 2.3368, 2.1153, 1.8011, 1.1771], + device='cuda:0'), covar=tensor([0.4791, 0.4637, 0.1476, 0.3103, 0.2172, 0.2774, 0.1928, 0.4737], + device='cuda:0'), in_proj_covar=tensor([0.0914, 0.0917, 0.0752, 0.0883, 0.0953, 0.0839, 0.0715, 0.0792], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 17:09:35,680 INFO [train.py:935] (0/4) Epoch 15, validation: loss=0.181, simple_loss=0.2808, pruned_loss=0.04056, over 944034.00 frames. +2023-02-06 17:09:35,681 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-06 17:09:37,104 INFO [optim.py:369] (0/4) 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 17:09:43,578 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 +2023-02-06 17:09:50,216 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.4373, 1.3351, 4.6758, 1.7576, 4.0698, 3.9052, 4.2198, 4.0681], + device='cuda:0'), covar=tensor([0.0668, 0.4574, 0.0444, 0.3724, 0.1079, 0.0834, 0.0568, 0.0675], + device='cuda:0'), in_proj_covar=tensor([0.0559, 0.0607, 0.0636, 0.0578, 0.0652, 0.0557, 0.0556, 0.0618], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 17:10:10,487 INFO [train.py:901] (0/4) Epoch 15, batch 6050, loss[loss=0.1879, simple_loss=0.2744, pruned_loss=0.05075, over 8032.00 frames. ], tot_loss[loss=0.222, simple_loss=0.3007, pruned_loss=0.07163, over 1616999.02 frames. ], batch size: 22, lr: 5.04e-03, grad_scale: 8.0 +2023-02-06 17:10:44,323 INFO [train.py:901] (0/4) Epoch 15, batch 6100, loss[loss=0.2984, simple_loss=0.3459, pruned_loss=0.1254, over 6561.00 frames. ], tot_loss[loss=0.2223, simple_loss=0.3009, pruned_loss=0.0718, over 1612521.93 frames. ], batch size: 71, lr: 5.04e-03, grad_scale: 8.0 +2023-02-06 17:10:45,658 INFO [optim.py:369] (0/4) 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 17:11:03,647 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 +2023-02-06 17:11:18,259 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-06 17:11:20,324 INFO [train.py:901] (0/4) Epoch 15, batch 6150, loss[loss=0.2859, simple_loss=0.3477, pruned_loss=0.112, over 6963.00 frames. ], tot_loss[loss=0.2216, simple_loss=0.3004, pruned_loss=0.07144, over 1612666.26 frames. ], batch size: 71, lr: 5.04e-03, grad_scale: 8.0 +2023-02-06 17:11:20,502 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.1845, 1.8088, 2.0668, 1.9518, 1.1490, 1.9164, 2.3780, 2.2223], + device='cuda:0'), covar=tensor([0.0400, 0.1144, 0.1553, 0.1220, 0.0607, 0.1328, 0.0605, 0.0531], + device='cuda:0'), in_proj_covar=tensor([0.0098, 0.0152, 0.0191, 0.0157, 0.0102, 0.0162, 0.0114, 0.0139], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 17:11:54,706 INFO [train.py:901] (0/4) Epoch 15, batch 6200, loss[loss=0.218, simple_loss=0.2854, pruned_loss=0.07529, over 7938.00 frames. ], tot_loss[loss=0.2213, simple_loss=0.3, pruned_loss=0.07131, over 1616890.52 frames. ], batch size: 20, lr: 5.04e-03, grad_scale: 8.0 +2023-02-06 17:11:55,624 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119364.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:11:56,081 INFO [optim.py:369] (0/4) 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 17:12:14,452 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119389.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:12:20,618 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0445, 1.2287, 1.1888, 0.6190, 1.2016, 1.0390, 0.0858, 1.1897], + device='cuda:0'), covar=tensor([0.0339, 0.0300, 0.0275, 0.0458, 0.0323, 0.0788, 0.0652, 0.0262], + device='cuda:0'), in_proj_covar=tensor([0.0423, 0.0364, 0.0315, 0.0422, 0.0350, 0.0512, 0.0378, 0.0391], + device='cuda:0'), out_proj_covar=tensor([1.1685e-04, 9.7884e-05, 8.4289e-05, 1.1401e-04, 9.4832e-05, 1.4925e-04, + 1.0413e-04, 1.0619e-04], device='cuda:0') +2023-02-06 17:12:24,803 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4512, 2.7333, 1.8234, 2.2318, 2.2795, 1.5887, 2.1006, 1.9897], + device='cuda:0'), covar=tensor([0.1406, 0.0299, 0.1084, 0.0612, 0.0631, 0.1340, 0.1031, 0.0867], + device='cuda:0'), in_proj_covar=tensor([0.0351, 0.0231, 0.0327, 0.0302, 0.0303, 0.0332, 0.0348, 0.0315], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 17:12:30,385 INFO [train.py:901] (0/4) Epoch 15, batch 6250, loss[loss=0.2453, simple_loss=0.3178, pruned_loss=0.08643, over 8123.00 frames. ], tot_loss[loss=0.221, simple_loss=0.2998, pruned_loss=0.07115, over 1618504.44 frames. ], batch size: 22, lr: 5.04e-03, grad_scale: 8.0 +2023-02-06 17:12:44,678 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8219, 1.5763, 3.1089, 1.4893, 2.3043, 3.3443, 3.4745, 2.6923], + device='cuda:0'), covar=tensor([0.1124, 0.1643, 0.0380, 0.2068, 0.0927, 0.0298, 0.0524, 0.0779], + device='cuda:0'), in_proj_covar=tensor([0.0276, 0.0305, 0.0269, 0.0298, 0.0285, 0.0248, 0.0374, 0.0297], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 17:12:47,158 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=119437.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:12:59,456 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=119455.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:13:04,851 INFO [train.py:901] (0/4) Epoch 15, batch 6300, loss[loss=0.2109, simple_loss=0.2891, pruned_loss=0.06636, over 8243.00 frames. ], tot_loss[loss=0.2204, simple_loss=0.299, pruned_loss=0.0709, over 1618851.24 frames. ], batch size: 22, lr: 5.04e-03, grad_scale: 8.0 +2023-02-06 17:13:06,142 INFO [optim.py:369] (0/4) 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 17:13:19,849 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.6754, 2.2500, 3.3548, 2.5465, 3.1130, 2.6176, 2.2245, 1.7076], + device='cuda:0'), covar=tensor([0.4436, 0.4842, 0.1559, 0.3240, 0.2321, 0.2429, 0.1684, 0.5412], + device='cuda:0'), in_proj_covar=tensor([0.0907, 0.0914, 0.0751, 0.0880, 0.0951, 0.0838, 0.0712, 0.0790], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 17:13:31,640 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0927, 1.2380, 1.2228, 0.7268, 1.2145, 0.9768, 0.2237, 1.1650], + device='cuda:0'), covar=tensor([0.0294, 0.0286, 0.0266, 0.0405, 0.0337, 0.0772, 0.0601, 0.0234], + device='cuda:0'), in_proj_covar=tensor([0.0418, 0.0361, 0.0313, 0.0419, 0.0347, 0.0506, 0.0375, 0.0386], + device='cuda:0'), out_proj_covar=tensor([1.1514e-04, 9.6950e-05, 8.3743e-05, 1.1325e-04, 9.3858e-05, 1.4731e-04, + 1.0318e-04, 1.0472e-04], device='cuda:0') +2023-02-06 17:13:41,043 INFO [train.py:901] (0/4) Epoch 15, batch 6350, loss[loss=0.2347, simple_loss=0.2922, pruned_loss=0.08857, over 7432.00 frames. ], tot_loss[loss=0.221, simple_loss=0.2991, pruned_loss=0.07141, over 1612360.54 frames. ], batch size: 17, lr: 5.03e-03, grad_scale: 8.0 +2023-02-06 17:13:53,784 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=119532.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:14:07,836 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=119552.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:14:15,092 INFO [train.py:901] (0/4) Epoch 15, batch 6400, loss[loss=0.193, simple_loss=0.2892, pruned_loss=0.0484, over 8104.00 frames. ], tot_loss[loss=0.2208, simple_loss=0.2991, pruned_loss=0.0713, over 1611708.89 frames. ], batch size: 23, lr: 5.03e-03, grad_scale: 8.0 +2023-02-06 17:14:16,455 INFO [optim.py:369] (0/4) 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 17:14:20,031 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=119570.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:14:41,458 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4201, 1.6438, 1.6902, 0.9120, 1.7132, 1.3068, 0.2555, 1.5866], + device='cuda:0'), covar=tensor([0.0333, 0.0246, 0.0245, 0.0395, 0.0278, 0.0733, 0.0648, 0.0188], + device='cuda:0'), in_proj_covar=tensor([0.0420, 0.0362, 0.0314, 0.0421, 0.0348, 0.0507, 0.0376, 0.0388], + device='cuda:0'), out_proj_covar=tensor([1.1572e-04, 9.7260e-05, 8.4003e-05, 1.1390e-04, 9.4212e-05, 1.4753e-04, + 1.0361e-04, 1.0534e-04], device='cuda:0') +2023-02-06 17:14:49,984 INFO [train.py:901] (0/4) Epoch 15, batch 6450, loss[loss=0.1823, simple_loss=0.2677, pruned_loss=0.0485, over 7808.00 frames. ], tot_loss[loss=0.2208, simple_loss=0.2991, pruned_loss=0.07129, over 1612517.64 frames. ], batch size: 19, lr: 5.03e-03, grad_scale: 8.0 +2023-02-06 17:15:24,240 INFO [train.py:901] (0/4) Epoch 15, batch 6500, loss[loss=0.2331, simple_loss=0.3063, pruned_loss=0.07994, over 8086.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.2988, pruned_loss=0.07116, over 1611042.30 frames. ], batch size: 21, lr: 5.03e-03, grad_scale: 8.0 +2023-02-06 17:15:25,570 INFO [optim.py:369] (0/4) 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 17:15:38,586 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=119683.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:15:58,716 INFO [train.py:901] (0/4) Epoch 15, batch 6550, loss[loss=0.2226, simple_loss=0.3002, pruned_loss=0.07246, over 8535.00 frames. ], tot_loss[loss=0.2202, simple_loss=0.2991, pruned_loss=0.0707, over 1614762.97 frames. ], batch size: 39, lr: 5.03e-03, grad_scale: 16.0 +2023-02-06 17:16:14,596 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1970, 1.9199, 2.6296, 2.1935, 2.3539, 2.2573, 1.8394, 1.2093], + device='cuda:0'), covar=tensor([0.4443, 0.4040, 0.1447, 0.2824, 0.2101, 0.2229, 0.1699, 0.4341], + device='cuda:0'), in_proj_covar=tensor([0.0909, 0.0918, 0.0756, 0.0881, 0.0953, 0.0841, 0.0716, 0.0794], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 17:16:29,721 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-06 17:16:34,447 INFO [train.py:901] (0/4) Epoch 15, batch 6600, loss[loss=0.2534, simple_loss=0.332, pruned_loss=0.08738, over 8459.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.2995, pruned_loss=0.07086, over 1608441.93 frames. ], batch size: 27, lr: 5.03e-03, grad_scale: 16.0 +2023-02-06 17:16:35,791 INFO [optim.py:369] (0/4) 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 17:16:48,573 WARNING [train.py:1067] (0/4) 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] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119808.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:17:08,621 INFO [train.py:901] (0/4) Epoch 15, batch 6650, loss[loss=0.2201, simple_loss=0.2968, pruned_loss=0.07171, over 7659.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.2997, pruned_loss=0.07124, over 1608445.50 frames. ], batch size: 19, lr: 5.03e-03, grad_scale: 16.0 +2023-02-06 17:17:17,667 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119826.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:17:22,300 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119833.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:17:36,358 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119851.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:17:44,390 INFO [train.py:901] (0/4) Epoch 15, batch 6700, loss[loss=0.2079, simple_loss=0.2891, pruned_loss=0.06336, over 8290.00 frames. ], tot_loss[loss=0.2198, simple_loss=0.2982, pruned_loss=0.07066, over 1607314.92 frames. ], batch size: 23, lr: 5.03e-03, grad_scale: 16.0 +2023-02-06 17:17:45,744 INFO [optim.py:369] (0/4) 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 17:17:46,267 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 +2023-02-06 17:17:48,208 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 +2023-02-06 17:17:53,411 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=119876.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:18:19,558 INFO [train.py:901] (0/4) Epoch 15, batch 6750, loss[loss=0.2128, simple_loss=0.2976, pruned_loss=0.06403, over 8334.00 frames. ], tot_loss[loss=0.2182, simple_loss=0.2966, pruned_loss=0.06989, over 1605457.60 frames. ], batch size: 26, lr: 5.03e-03, grad_scale: 16.0 +2023-02-06 17:18:55,246 INFO [train.py:901] (0/4) Epoch 15, batch 6800, loss[loss=0.2127, simple_loss=0.2949, pruned_loss=0.06521, over 8345.00 frames. ], tot_loss[loss=0.2182, simple_loss=0.2972, pruned_loss=0.06957, over 1611359.03 frames. ], batch size: 26, lr: 5.02e-03, grad_scale: 16.0 +2023-02-06 17:18:57,360 INFO [optim.py:369] (0/4) 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 17:19:03,578 WARNING [train.py:1067] (0/4) 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] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=119991.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:19:21,553 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-120000.pt +2023-02-06 17:19:32,107 INFO [train.py:901] (0/4) Epoch 15, batch 6850, loss[loss=0.339, simple_loss=0.3796, pruned_loss=0.1492, over 7059.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.2973, pruned_loss=0.06947, over 1610968.62 frames. ], batch size: 73, lr: 5.02e-03, grad_scale: 16.0 +2023-02-06 17:19:33,898 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-02-06 17:19:41,642 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=120027.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:19:46,726 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2558, 1.9665, 2.6921, 2.2388, 2.6007, 2.2042, 1.8274, 1.4106], + device='cuda:0'), covar=tensor([0.4775, 0.4364, 0.1509, 0.2956, 0.2043, 0.2514, 0.1785, 0.4400], + device='cuda:0'), in_proj_covar=tensor([0.0911, 0.0921, 0.0757, 0.0881, 0.0952, 0.0843, 0.0718, 0.0796], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 17:19:53,401 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-06 17:20:06,209 INFO [train.py:901] (0/4) Epoch 15, batch 6900, loss[loss=0.2107, simple_loss=0.2915, pruned_loss=0.06492, over 7199.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.2972, pruned_loss=0.06952, over 1611507.73 frames. ], batch size: 16, lr: 5.02e-03, grad_scale: 16.0 +2023-02-06 17:20:07,530 INFO [optim.py:369] (0/4) 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 17:20:23,943 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0027, 1.3518, 1.6023, 1.2194, 0.9167, 1.4375, 1.5317, 1.4447], + device='cuda:0'), covar=tensor([0.0478, 0.1269, 0.1737, 0.1482, 0.0631, 0.1488, 0.0743, 0.0635], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0153, 0.0190, 0.0156, 0.0100, 0.0163, 0.0113, 0.0139], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 17:20:42,260 INFO [train.py:901] (0/4) Epoch 15, batch 6950, loss[loss=0.213, simple_loss=0.3044, pruned_loss=0.06081, over 8335.00 frames. ], tot_loss[loss=0.218, simple_loss=0.2973, pruned_loss=0.06937, over 1613528.33 frames. ], batch size: 25, lr: 5.02e-03, grad_scale: 16.0 +2023-02-06 17:21:02,380 INFO [zipformer.py:1185] (0/4) 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] (0/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-06 17:21:11,749 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.9938, 1.6676, 1.8379, 1.6229, 1.0726, 1.6523, 2.3554, 2.1642], + device='cuda:0'), covar=tensor([0.0404, 0.1282, 0.1703, 0.1389, 0.0609, 0.1546, 0.0600, 0.0558], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0153, 0.0190, 0.0156, 0.0100, 0.0162, 0.0114, 0.0139], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 17:21:11,999 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.96 vs. limit=5.0 +2023-02-06 17:21:16,260 INFO [train.py:901] (0/4) Epoch 15, batch 7000, loss[loss=0.2021, simple_loss=0.2785, pruned_loss=0.06286, over 7547.00 frames. ], tot_loss[loss=0.2176, simple_loss=0.2969, pruned_loss=0.06916, over 1612649.35 frames. ], batch size: 18, lr: 5.02e-03, grad_scale: 16.0 +2023-02-06 17:21:17,612 INFO [optim.py:369] (0/4) 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 17:21:22,337 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.4404, 4.3388, 3.9310, 2.0128, 3.9478, 3.9541, 4.0647, 3.7054], + device='cuda:0'), covar=tensor([0.0696, 0.0510, 0.0896, 0.4494, 0.0813, 0.1067, 0.1124, 0.0998], + device='cuda:0'), in_proj_covar=tensor([0.0497, 0.0409, 0.0417, 0.0513, 0.0408, 0.0415, 0.0397, 0.0361], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 17:21:51,897 INFO [train.py:901] (0/4) Epoch 15, batch 7050, loss[loss=0.2232, simple_loss=0.2811, pruned_loss=0.08262, over 7706.00 frames. ], tot_loss[loss=0.2177, simple_loss=0.2966, pruned_loss=0.06936, over 1609648.84 frames. ], batch size: 18, lr: 5.02e-03, grad_scale: 16.0 +2023-02-06 17:22:15,018 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=120247.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:22:26,166 INFO [train.py:901] (0/4) Epoch 15, batch 7100, loss[loss=0.2259, simple_loss=0.3036, pruned_loss=0.07406, over 8098.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.296, pruned_loss=0.06884, over 1610670.56 frames. ], batch size: 23, lr: 5.02e-03, grad_scale: 16.0 +2023-02-06 17:22:27,488 INFO [optim.py:369] (0/4) 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 17:22:32,360 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=120272.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:22:35,797 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4735, 1.9832, 3.3072, 1.3303, 2.3614, 1.9169, 1.5980, 2.2914], + device='cuda:0'), covar=tensor([0.1783, 0.2099, 0.0743, 0.4065, 0.1602, 0.2810, 0.1949, 0.2195], + device='cuda:0'), in_proj_covar=tensor([0.0500, 0.0555, 0.0538, 0.0606, 0.0628, 0.0569, 0.0498, 0.0622], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 17:22:58,518 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.92 vs. limit=5.0 +2023-02-06 17:23:00,845 INFO [train.py:901] (0/4) Epoch 15, batch 7150, loss[loss=0.2064, simple_loss=0.2881, pruned_loss=0.06238, over 8299.00 frames. ], tot_loss[loss=0.216, simple_loss=0.2952, pruned_loss=0.06843, over 1611101.15 frames. ], batch size: 48, lr: 5.02e-03, grad_scale: 16.0 +2023-02-06 17:23:35,467 INFO [train.py:901] (0/4) Epoch 15, batch 7200, loss[loss=0.239, simple_loss=0.3222, pruned_loss=0.07789, over 8353.00 frames. ], tot_loss[loss=0.217, simple_loss=0.2958, pruned_loss=0.06907, over 1609841.11 frames. ], batch size: 24, lr: 5.02e-03, grad_scale: 16.0 +2023-02-06 17:23:36,812 INFO [optim.py:369] (0/4) 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 17:23:49,344 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.6281, 4.6683, 4.0880, 1.8974, 4.1337, 4.1933, 4.2429, 3.9713], + device='cuda:0'), covar=tensor([0.0710, 0.0555, 0.1174, 0.5454, 0.0875, 0.1095, 0.1313, 0.0955], + device='cuda:0'), in_proj_covar=tensor([0.0498, 0.0415, 0.0418, 0.0518, 0.0410, 0.0420, 0.0403, 0.0364], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 17:24:00,219 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=120398.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:24:10,194 INFO [train.py:901] (0/4) Epoch 15, batch 7250, loss[loss=0.241, simple_loss=0.3176, pruned_loss=0.08225, over 8639.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.2942, pruned_loss=0.06811, over 1604836.51 frames. ], batch size: 34, lr: 5.02e-03, grad_scale: 16.0 +2023-02-06 17:24:17,873 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=120423.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:24:45,987 INFO [train.py:901] (0/4) Epoch 15, batch 7300, loss[loss=0.2265, simple_loss=0.3026, pruned_loss=0.07518, over 8713.00 frames. ], tot_loss[loss=0.2171, simple_loss=0.2958, pruned_loss=0.06916, over 1605002.36 frames. ], batch size: 34, lr: 5.01e-03, grad_scale: 16.0 +2023-02-06 17:24:47,342 INFO [optim.py:369] (0/4) 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 17:25:20,537 INFO [train.py:901] (0/4) Epoch 15, batch 7350, loss[loss=0.164, simple_loss=0.2478, pruned_loss=0.04008, over 7202.00 frames. ], tot_loss[loss=0.2177, simple_loss=0.2966, pruned_loss=0.06937, over 1607857.78 frames. ], batch size: 16, lr: 5.01e-03, grad_scale: 16.0 +2023-02-06 17:25:45,388 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-06 17:25:55,000 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9143, 1.3112, 1.5604, 1.1643, 0.8263, 1.3578, 1.5085, 1.2897], + device='cuda:0'), covar=tensor([0.0498, 0.1324, 0.1778, 0.1534, 0.0634, 0.1557, 0.0757, 0.0714], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0152, 0.0190, 0.0157, 0.0100, 0.0162, 0.0113, 0.0139], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 17:25:56,239 INFO [train.py:901] (0/4) Epoch 15, batch 7400, loss[loss=0.2028, simple_loss=0.2917, pruned_loss=0.05701, over 8317.00 frames. ], tot_loss[loss=0.218, simple_loss=0.2968, pruned_loss=0.06956, over 1608228.91 frames. ], batch size: 25, lr: 5.01e-03, grad_scale: 16.0 +2023-02-06 17:25:57,543 INFO [optim.py:369] (0/4) 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 17:26:04,618 WARNING [train.py:1067] (0/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-06 17:26:30,892 INFO [train.py:901] (0/4) Epoch 15, batch 7450, loss[loss=0.1743, simple_loss=0.2618, pruned_loss=0.04338, over 8093.00 frames. ], tot_loss[loss=0.2203, simple_loss=0.2989, pruned_loss=0.07084, over 1609751.10 frames. ], batch size: 21, lr: 5.01e-03, grad_scale: 16.0 +2023-02-06 17:26:42,795 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 15, batch 7500, loss[loss=0.2228, simple_loss=0.3084, pruned_loss=0.06866, over 8337.00 frames. ], tot_loss[loss=0.2201, simple_loss=0.2991, pruned_loss=0.07054, over 1615141.29 frames. ], batch size: 26, lr: 5.01e-03, grad_scale: 16.0 +2023-02-06 17:27:07,304 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2495, 2.5006, 3.0272, 1.5478, 3.2018, 1.7365, 1.4362, 2.1872], + device='cuda:0'), covar=tensor([0.0721, 0.0340, 0.0248, 0.0692, 0.0474, 0.0776, 0.0900, 0.0413], + device='cuda:0'), in_proj_covar=tensor([0.0418, 0.0362, 0.0312, 0.0416, 0.0347, 0.0503, 0.0369, 0.0384], + device='cuda:0'), out_proj_covar=tensor([1.1535e-04, 9.7416e-05, 8.3435e-05, 1.1223e-04, 9.3515e-05, 1.4610e-04, + 1.0167e-04, 1.0432e-04], device='cuda:0') +2023-02-06 17:27:07,754 INFO [optim.py:369] (0/4) 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 17:27:19,906 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9433, 1.6366, 2.2550, 1.8541, 2.1027, 1.9124, 1.6162, 1.0479], + device='cuda:0'), covar=tensor([0.4326, 0.3728, 0.1272, 0.2504, 0.1813, 0.2295, 0.1713, 0.3831], + device='cuda:0'), in_proj_covar=tensor([0.0909, 0.0918, 0.0755, 0.0881, 0.0951, 0.0841, 0.0717, 0.0793], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 17:27:27,351 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=120694.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:27:30,018 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.8173, 1.2724, 3.9527, 1.3583, 3.4951, 3.2755, 3.5670, 3.4568], + device='cuda:0'), covar=tensor([0.0583, 0.4594, 0.0612, 0.4059, 0.1128, 0.0984, 0.0637, 0.0741], + device='cuda:0'), in_proj_covar=tensor([0.0564, 0.0616, 0.0639, 0.0590, 0.0662, 0.0568, 0.0560, 0.0621], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 17:27:40,240 INFO [train.py:901] (0/4) Epoch 15, batch 7550, loss[loss=0.2283, simple_loss=0.3073, pruned_loss=0.07461, over 7666.00 frames. ], tot_loss[loss=0.2197, simple_loss=0.2987, pruned_loss=0.07032, over 1617649.65 frames. ], batch size: 19, lr: 5.01e-03, grad_scale: 16.0 +2023-02-06 17:28:14,839 INFO [train.py:901] (0/4) Epoch 15, batch 7600, loss[loss=0.2198, simple_loss=0.303, pruned_loss=0.06825, over 8138.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.2981, pruned_loss=0.07036, over 1616282.36 frames. ], batch size: 22, lr: 5.01e-03, grad_scale: 16.0 +2023-02-06 17:28:16,206 INFO [optim.py:369] (0/4) 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 17:28:22,209 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 +2023-02-06 17:28:46,876 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.1819, 1.0535, 1.2612, 1.0858, 0.9084, 1.2959, 0.0414, 0.9344], + device='cuda:0'), covar=tensor([0.1758, 0.1590, 0.0519, 0.0964, 0.3260, 0.0558, 0.2494, 0.1403], + device='cuda:0'), in_proj_covar=tensor([0.0172, 0.0179, 0.0111, 0.0214, 0.0257, 0.0115, 0.0162, 0.0176], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 17:28:50,145 INFO [train.py:901] (0/4) Epoch 15, batch 7650, loss[loss=0.2072, simple_loss=0.2922, pruned_loss=0.06113, over 8294.00 frames. ], tot_loss[loss=0.2199, simple_loss=0.2985, pruned_loss=0.07066, over 1617165.55 frames. ], batch size: 23, lr: 5.01e-03, grad_scale: 8.0 +2023-02-06 17:29:23,492 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6444, 1.5071, 3.0722, 1.2131, 2.1323, 3.3763, 3.4378, 2.9010], + device='cuda:0'), covar=tensor([0.1268, 0.1767, 0.0406, 0.2308, 0.1113, 0.0265, 0.0632, 0.0596], + device='cuda:0'), in_proj_covar=tensor([0.0279, 0.0308, 0.0272, 0.0301, 0.0287, 0.0249, 0.0377, 0.0297], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 17:29:25,336 INFO [train.py:901] (0/4) Epoch 15, batch 7700, loss[loss=0.2735, simple_loss=0.345, pruned_loss=0.101, over 8475.00 frames. ], tot_loss[loss=0.2198, simple_loss=0.298, pruned_loss=0.07074, over 1612082.51 frames. ], batch size: 25, lr: 5.01e-03, grad_scale: 8.0 +2023-02-06 17:29:27,396 INFO [optim.py:369] (0/4) 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 17:29:35,157 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.0304, 2.6953, 3.6937, 1.7898, 1.7975, 3.6356, 0.8148, 2.0880], + device='cuda:0'), covar=tensor([0.1946, 0.1305, 0.0292, 0.2205, 0.3522, 0.0304, 0.2590, 0.1835], + device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0181, 0.0112, 0.0215, 0.0259, 0.0116, 0.0162, 0.0177], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 17:29:52,766 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 15, batch 7750, loss[loss=0.2145, simple_loss=0.3017, pruned_loss=0.06371, over 8197.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.2978, pruned_loss=0.07051, over 1609347.20 frames. ], batch size: 23, lr: 5.01e-03, grad_scale: 8.0 +2023-02-06 17:30:36,084 INFO [train.py:901] (0/4) Epoch 15, batch 7800, loss[loss=0.2325, simple_loss=0.3332, pruned_loss=0.06595, over 8279.00 frames. ], tot_loss[loss=0.2188, simple_loss=0.2972, pruned_loss=0.07016, over 1607872.13 frames. ], batch size: 23, lr: 5.00e-03, grad_scale: 8.0 +2023-02-06 17:30:38,106 INFO [optim.py:369] (0/4) 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 17:31:09,475 INFO [train.py:901] (0/4) Epoch 15, batch 7850, loss[loss=0.2099, simple_loss=0.2978, pruned_loss=0.061, over 8553.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.2987, pruned_loss=0.07123, over 1609392.29 frames. ], batch size: 49, lr: 5.00e-03, grad_scale: 8.0 +2023-02-06 17:31:14,868 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121021.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:31:26,046 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=121038.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:31:42,564 INFO [train.py:901] (0/4) Epoch 15, batch 7900, loss[loss=0.2245, simple_loss=0.3091, pruned_loss=0.06989, over 8332.00 frames. ], tot_loss[loss=0.2198, simple_loss=0.2985, pruned_loss=0.07054, over 1614873.46 frames. ], batch size: 25, lr: 5.00e-03, grad_scale: 8.0 +2023-02-06 17:31:44,516 INFO [optim.py:369] (0/4) 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 17:32:15,970 INFO [train.py:901] (0/4) Epoch 15, batch 7950, loss[loss=0.2373, simple_loss=0.3205, pruned_loss=0.07703, over 8457.00 frames. ], tot_loss[loss=0.2204, simple_loss=0.2995, pruned_loss=0.07065, over 1618225.71 frames. ], batch size: 25, lr: 5.00e-03, grad_scale: 8.0 +2023-02-06 17:32:42,027 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121153.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:32:48,354 INFO [train.py:901] (0/4) Epoch 15, batch 8000, loss[loss=0.2351, simple_loss=0.317, pruned_loss=0.07662, over 8511.00 frames. ], tot_loss[loss=0.2219, simple_loss=0.3005, pruned_loss=0.07167, over 1618465.26 frames. ], batch size: 28, lr: 5.00e-03, grad_scale: 8.0 +2023-02-06 17:32:50,387 INFO [optim.py:369] (0/4) 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 17:33:01,506 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121182.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:33:22,841 INFO [train.py:901] (0/4) Epoch 15, batch 8050, loss[loss=0.1895, simple_loss=0.2661, pruned_loss=0.05644, over 7522.00 frames. ], tot_loss[loss=0.221, simple_loss=0.2989, pruned_loss=0.07151, over 1599696.96 frames. ], batch size: 18, lr: 5.00e-03, grad_scale: 8.0 +2023-02-06 17:33:46,150 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/epoch-15.pt +2023-02-06 17:33:57,581 WARNING [train.py:1067] (0/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-06 17:34:00,918 INFO [train.py:901] (0/4) Epoch 16, batch 0, loss[loss=0.2319, simple_loss=0.3179, pruned_loss=0.07293, over 8515.00 frames. ], tot_loss[loss=0.2319, simple_loss=0.3179, pruned_loss=0.07293, over 8515.00 frames. ], batch size: 28, lr: 4.84e-03, grad_scale: 8.0 +2023-02-06 17:34:00,919 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 17:34:11,306 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5537, 1.5936, 2.7108, 1.2640, 2.0183, 2.8857, 3.0769, 2.4349], + device='cuda:0'), covar=tensor([0.1385, 0.1671, 0.0440, 0.2557, 0.0941, 0.0394, 0.0527, 0.0856], + device='cuda:0'), in_proj_covar=tensor([0.0280, 0.0309, 0.0271, 0.0301, 0.0288, 0.0248, 0.0377, 0.0297], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 17:34:11,910 INFO [train.py:935] (0/4) Epoch 16, validation: loss=0.1795, simple_loss=0.2801, pruned_loss=0.03944, over 944034.00 frames. +2023-02-06 17:34:11,911 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-06 17:34:24,910 INFO [optim.py:369] (0/4) 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] (0/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-06 17:34:47,413 INFO [train.py:901] (0/4) Epoch 16, batch 50, loss[loss=0.2304, simple_loss=0.3232, pruned_loss=0.06878, over 8463.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.295, pruned_loss=0.06775, over 361350.08 frames. ], batch size: 25, lr: 4.84e-03, grad_scale: 8.0 +2023-02-06 17:35:02,273 WARNING [train.py:1067] (0/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-06 17:35:02,487 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4507, 2.8246, 3.4042, 1.8475, 3.4849, 2.2340, 1.6268, 2.2199], + device='cuda:0'), covar=tensor([0.0695, 0.0334, 0.0168, 0.0586, 0.0348, 0.0569, 0.0740, 0.0407], + device='cuda:0'), in_proj_covar=tensor([0.0417, 0.0359, 0.0311, 0.0416, 0.0347, 0.0506, 0.0368, 0.0381], + device='cuda:0'), out_proj_covar=tensor([1.1499e-04, 9.6322e-05, 8.3087e-05, 1.1218e-04, 9.3558e-05, 1.4700e-04, + 1.0127e-04, 1.0336e-04], device='cuda:0') +2023-02-06 17:35:09,786 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121329.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 17:35:20,723 INFO [train.py:901] (0/4) Epoch 16, batch 100, loss[loss=0.2039, simple_loss=0.2929, pruned_loss=0.05743, over 8090.00 frames. ], tot_loss[loss=0.2208, simple_loss=0.3, pruned_loss=0.07083, over 638816.16 frames. ], batch size: 21, lr: 4.84e-03, grad_scale: 8.0 +2023-02-06 17:35:24,727 WARNING [train.py:1067] (0/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-06 17:35:33,276 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 17:35:42,089 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4043, 2.0283, 2.8034, 2.2044, 2.6743, 2.3195, 2.0680, 1.4485], + device='cuda:0'), covar=tensor([0.4476, 0.4258, 0.1482, 0.3589, 0.2354, 0.2508, 0.1695, 0.4683], + device='cuda:0'), in_proj_covar=tensor([0.0906, 0.0914, 0.0751, 0.0883, 0.0946, 0.0837, 0.0717, 0.0793], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 17:35:53,848 INFO [train.py:901] (0/4) Epoch 16, batch 150, loss[loss=0.2509, simple_loss=0.3223, pruned_loss=0.08982, over 8590.00 frames. ], tot_loss[loss=0.2216, simple_loss=0.3002, pruned_loss=0.07152, over 856532.45 frames. ], batch size: 31, lr: 4.84e-03, grad_scale: 8.0 +2023-02-06 17:36:04,295 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=121409.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:36:15,647 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121425.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:36:21,671 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121434.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:36:30,075 INFO [train.py:901] (0/4) Epoch 16, batch 200, loss[loss=0.1758, simple_loss=0.247, pruned_loss=0.05231, over 7404.00 frames. ], tot_loss[loss=0.223, simple_loss=0.3009, pruned_loss=0.07254, over 1022029.11 frames. ], batch size: 17, lr: 4.83e-03, grad_scale: 8.0 +2023-02-06 17:36:43,673 INFO [optim.py:369] (0/4) 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 17:36:46,560 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6446, 1.9188, 2.0827, 1.4042, 2.1879, 1.5448, 0.5543, 1.8716], + device='cuda:0'), covar=tensor([0.0478, 0.0282, 0.0187, 0.0388, 0.0294, 0.0655, 0.0670, 0.0219], + device='cuda:0'), in_proj_covar=tensor([0.0415, 0.0359, 0.0310, 0.0415, 0.0346, 0.0504, 0.0367, 0.0380], + device='cuda:0'), out_proj_covar=tensor([1.1452e-04, 9.6256e-05, 8.2914e-05, 1.1203e-04, 9.3313e-05, 1.4621e-04, + 1.0085e-04, 1.0301e-04], device='cuda:0') +2023-02-06 17:36:51,926 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.7148, 5.7860, 5.1142, 2.1942, 5.1172, 5.4380, 5.3785, 5.2022], + device='cuda:0'), covar=tensor([0.0591, 0.0423, 0.0848, 0.4711, 0.0713, 0.0678, 0.1029, 0.0646], + device='cuda:0'), in_proj_covar=tensor([0.0488, 0.0404, 0.0409, 0.0512, 0.0401, 0.0407, 0.0392, 0.0355], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 17:36:53,338 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121480.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:37:04,017 INFO [train.py:901] (0/4) Epoch 16, batch 250, loss[loss=0.2211, simple_loss=0.3051, pruned_loss=0.06851, over 8514.00 frames. ], tot_loss[loss=0.2218, simple_loss=0.3001, pruned_loss=0.07175, over 1150620.35 frames. ], batch size: 39, lr: 4.83e-03, grad_scale: 8.0 +2023-02-06 17:37:18,656 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-06 17:37:24,805 INFO [zipformer.py:1185] (0/4) 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] (0/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-06 17:37:30,229 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.3923, 4.3772, 3.9741, 1.9577, 3.9360, 3.8672, 3.9270, 3.7169], + device='cuda:0'), covar=tensor([0.0733, 0.0563, 0.0957, 0.4496, 0.0821, 0.1148, 0.1277, 0.0956], + device='cuda:0'), in_proj_covar=tensor([0.0493, 0.0409, 0.0412, 0.0516, 0.0404, 0.0411, 0.0398, 0.0358], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 17:37:39,709 INFO [train.py:901] (0/4) Epoch 16, batch 300, loss[loss=0.2191, simple_loss=0.2981, pruned_loss=0.07011, over 8577.00 frames. ], tot_loss[loss=0.2212, simple_loss=0.2997, pruned_loss=0.07137, over 1255926.41 frames. ], batch size: 39, lr: 4.83e-03, grad_scale: 8.0 +2023-02-06 17:37:50,101 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7517, 1.6157, 2.3651, 1.6366, 1.2619, 2.4131, 0.2837, 1.3375], + device='cuda:0'), covar=tensor([0.1993, 0.1603, 0.0370, 0.1640, 0.3550, 0.0447, 0.3109, 0.1773], + device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0180, 0.0111, 0.0214, 0.0256, 0.0115, 0.0162, 0.0177], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 17:37:54,075 INFO [optim.py:369] (0/4) 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 17:38:14,590 INFO [train.py:901] (0/4) Epoch 16, batch 350, loss[loss=0.2139, simple_loss=0.2772, pruned_loss=0.0753, over 7786.00 frames. ], tot_loss[loss=0.221, simple_loss=0.2994, pruned_loss=0.07126, over 1332449.97 frames. ], batch size: 19, lr: 4.83e-03, grad_scale: 8.0 +2023-02-06 17:38:45,986 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121641.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:38:49,787 INFO [train.py:901] (0/4) Epoch 16, batch 400, loss[loss=0.2203, simple_loss=0.3023, pruned_loss=0.06916, over 8504.00 frames. ], tot_loss[loss=0.2218, simple_loss=0.3, pruned_loss=0.07175, over 1396656.80 frames. ], batch size: 26, lr: 4.83e-03, grad_scale: 8.0 +2023-02-06 17:39:04,288 INFO [optim.py:369] (0/4) 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 17:39:09,177 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=121673.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 17:39:25,134 INFO [train.py:901] (0/4) Epoch 16, batch 450, loss[loss=0.2388, simple_loss=0.3324, pruned_loss=0.07259, over 8460.00 frames. ], tot_loss[loss=0.2219, simple_loss=0.301, pruned_loss=0.07141, over 1447582.57 frames. ], batch size: 27, lr: 4.83e-03, grad_scale: 8.0 +2023-02-06 17:39:52,393 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=121736.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:39:59,040 INFO [train.py:901] (0/4) Epoch 16, batch 500, loss[loss=0.235, simple_loss=0.312, pruned_loss=0.07906, over 8100.00 frames. ], tot_loss[loss=0.2225, simple_loss=0.3013, pruned_loss=0.07189, over 1485022.81 frames. ], batch size: 23, lr: 4.83e-03, grad_scale: 8.0 +2023-02-06 17:40:00,112 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.39 vs. limit=5.0 +2023-02-06 17:40:10,960 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121761.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:40:14,771 INFO [optim.py:369] (0/4) 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 17:40:17,014 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121788.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 17:40:35,814 INFO [train.py:901] (0/4) Epoch 16, batch 550, loss[loss=0.2389, simple_loss=0.319, pruned_loss=0.07936, over 8334.00 frames. ], tot_loss[loss=0.2218, simple_loss=0.3009, pruned_loss=0.07138, over 1512640.63 frames. ], batch size: 26, lr: 4.83e-03, grad_scale: 8.0 +2023-02-06 17:41:02,682 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0241, 1.5392, 3.4373, 1.4517, 2.3245, 3.7796, 3.8938, 3.2391], + device='cuda:0'), covar=tensor([0.1113, 0.1798, 0.0302, 0.2130, 0.1073, 0.0225, 0.0367, 0.0587], + device='cuda:0'), in_proj_covar=tensor([0.0281, 0.0310, 0.0273, 0.0301, 0.0292, 0.0248, 0.0381, 0.0298], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 17:41:09,179 INFO [train.py:901] (0/4) Epoch 16, batch 600, loss[loss=0.1837, simple_loss=0.2663, pruned_loss=0.05053, over 8224.00 frames. ], tot_loss[loss=0.2209, simple_loss=0.2998, pruned_loss=0.07099, over 1535264.28 frames. ], batch size: 22, lr: 4.83e-03, grad_scale: 8.0 +2023-02-06 17:41:15,570 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 +2023-02-06 17:41:22,438 INFO [optim.py:369] (0/4) 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 17:41:26,598 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-06 17:41:32,816 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6739, 2.2304, 4.2757, 1.4942, 2.9973, 2.2262, 1.7156, 2.9005], + device='cuda:0'), covar=tensor([0.1815, 0.2598, 0.0682, 0.4215, 0.1674, 0.2991, 0.2138, 0.2376], + device='cuda:0'), in_proj_covar=tensor([0.0504, 0.0556, 0.0538, 0.0608, 0.0629, 0.0576, 0.0501, 0.0622], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 17:41:36,811 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 16, batch 650, loss[loss=0.227, simple_loss=0.3023, pruned_loss=0.07586, over 8335.00 frames. ], tot_loss[loss=0.2198, simple_loss=0.2987, pruned_loss=0.07048, over 1549211.09 frames. ], batch size: 26, lr: 4.83e-03, grad_scale: 8.0 +2023-02-06 17:41:45,640 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=121897.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:41:46,947 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6289, 2.1379, 3.4091, 1.4015, 2.4042, 1.9613, 1.7804, 2.2099], + device='cuda:0'), covar=tensor([0.1635, 0.1994, 0.0666, 0.3947, 0.1623, 0.2836, 0.1821, 0.2265], + device='cuda:0'), in_proj_covar=tensor([0.0504, 0.0555, 0.0538, 0.0608, 0.0629, 0.0575, 0.0500, 0.0621], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 17:42:02,840 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121922.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:42:03,506 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0156, 2.2797, 1.8336, 2.7313, 1.3619, 1.6323, 1.8956, 2.2428], + device='cuda:0'), covar=tensor([0.0737, 0.0764, 0.0952, 0.0377, 0.1137, 0.1326, 0.0941, 0.0776], + device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0204, 0.0251, 0.0214, 0.0215, 0.0251, 0.0256, 0.0216], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 17:42:18,672 INFO [train.py:901] (0/4) Epoch 16, batch 700, loss[loss=0.2169, simple_loss=0.3086, pruned_loss=0.06264, over 8258.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.2976, pruned_loss=0.06954, over 1558923.19 frames. ], batch size: 24, lr: 4.82e-03, grad_scale: 8.0 +2023-02-06 17:42:32,105 INFO [optim.py:369] (0/4) 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 17:42:33,590 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121968.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:42:53,714 INFO [train.py:901] (0/4) Epoch 16, batch 750, loss[loss=0.2327, simple_loss=0.3054, pruned_loss=0.07997, over 8474.00 frames. ], tot_loss[loss=0.218, simple_loss=0.2973, pruned_loss=0.06938, over 1570236.11 frames. ], batch size: 29, lr: 4.82e-03, grad_scale: 8.0 +2023-02-06 17:42:56,595 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-122000.pt +2023-02-06 17:43:02,798 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7232, 1.4951, 2.8164, 1.3233, 2.0619, 3.0301, 3.1816, 2.5757], + device='cuda:0'), covar=tensor([0.1067, 0.1502, 0.0370, 0.2085, 0.0952, 0.0293, 0.0644, 0.0623], + device='cuda:0'), in_proj_covar=tensor([0.0280, 0.0308, 0.0272, 0.0299, 0.0291, 0.0248, 0.0379, 0.0296], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-06 17:43:14,285 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-06 17:43:23,876 WARNING [train.py:1067] (0/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-06 17:43:28,681 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 16, batch 800, loss[loss=0.2002, simple_loss=0.2768, pruned_loss=0.06182, over 7817.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.2986, pruned_loss=0.07013, over 1581664.45 frames. ], batch size: 20, lr: 4.82e-03, grad_scale: 8.0 +2023-02-06 17:43:43,080 INFO [optim.py:369] (0/4) 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] (0/4) 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] (0/4) Epoch 16, batch 850, loss[loss=0.2173, simple_loss=0.2986, pruned_loss=0.06801, over 8306.00 frames. ], tot_loss[loss=0.2184, simple_loss=0.2981, pruned_loss=0.06933, over 1595186.80 frames. ], batch size: 23, lr: 4.82e-03, grad_scale: 8.0 +2023-02-06 17:44:31,619 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122135.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:44:34,936 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122140.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:44:38,755 INFO [train.py:901] (0/4) Epoch 16, batch 900, loss[loss=0.2122, simple_loss=0.2939, pruned_loss=0.06528, over 8138.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.2991, pruned_loss=0.06984, over 1600963.30 frames. ], batch size: 22, lr: 4.82e-03, grad_scale: 8.0 +2023-02-06 17:44:52,346 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122165.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:44:52,800 INFO [optim.py:369] (0/4) 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 17:45:12,889 INFO [train.py:901] (0/4) Epoch 16, batch 950, loss[loss=0.2632, simple_loss=0.3148, pruned_loss=0.1058, over 5947.00 frames. ], tot_loss[loss=0.221, simple_loss=0.3, pruned_loss=0.07104, over 1600676.60 frames. ], batch size: 13, lr: 4.82e-03, grad_scale: 8.0 +2023-02-06 17:45:24,847 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122213.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:45:40,115 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-06 17:45:49,037 INFO [train.py:901] (0/4) Epoch 16, batch 1000, loss[loss=0.2343, simple_loss=0.3113, pruned_loss=0.07864, over 8197.00 frames. ], tot_loss[loss=0.2195, simple_loss=0.2987, pruned_loss=0.07013, over 1603702.22 frames. ], batch size: 23, lr: 4.82e-03, grad_scale: 8.0 +2023-02-06 17:46:03,407 INFO [optim.py:369] (0/4) 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 17:46:14,173 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-06 17:46:23,697 INFO [train.py:901] (0/4) Epoch 16, batch 1050, loss[loss=0.2286, simple_loss=0.3127, pruned_loss=0.07227, over 8361.00 frames. ], tot_loss[loss=0.2203, simple_loss=0.2993, pruned_loss=0.07062, over 1609348.27 frames. ], batch size: 24, lr: 4.82e-03, grad_scale: 8.0 +2023-02-06 17:46:26,430 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-06 17:46:32,121 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6844, 1.6333, 2.0824, 1.5515, 1.2096, 2.0929, 0.1828, 1.2292], + device='cuda:0'), covar=tensor([0.2021, 0.1348, 0.0460, 0.1293, 0.3350, 0.0466, 0.2770, 0.1714], + device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0180, 0.0112, 0.0211, 0.0255, 0.0116, 0.0161, 0.0176], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 17:46:34,622 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=122312.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:46:40,811 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3300, 1.5377, 2.0370, 1.2432, 1.3850, 1.5876, 1.4336, 1.4268], + device='cuda:0'), covar=tensor([0.1767, 0.2318, 0.0898, 0.3996, 0.1754, 0.2954, 0.2010, 0.2096], + device='cuda:0'), in_proj_covar=tensor([0.0503, 0.0555, 0.0538, 0.0605, 0.0625, 0.0569, 0.0499, 0.0619], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 17:46:51,297 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 16, batch 1100, loss[loss=0.1954, simple_loss=0.2868, pruned_loss=0.05202, over 8474.00 frames. ], tot_loss[loss=0.219, simple_loss=0.2982, pruned_loss=0.06996, over 1612292.86 frames. ], batch size: 25, lr: 4.82e-03, grad_scale: 8.0 +2023-02-06 17:47:04,883 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 +2023-02-06 17:47:12,623 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 16, batch 1150, loss[loss=0.2314, simple_loss=0.3142, pruned_loss=0.07435, over 8469.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.2985, pruned_loss=0.06987, over 1617763.03 frames. ], batch size: 27, lr: 4.82e-03, grad_scale: 8.0 +2023-02-06 17:47:38,306 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-06 17:47:54,847 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122427.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:47:58,515 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 +2023-02-06 17:48:07,557 INFO [train.py:901] (0/4) Epoch 16, batch 1200, loss[loss=0.2147, simple_loss=0.2903, pruned_loss=0.06959, over 7920.00 frames. ], tot_loss[loss=0.219, simple_loss=0.2982, pruned_loss=0.06986, over 1616324.22 frames. ], batch size: 20, lr: 4.81e-03, grad_scale: 8.0 +2023-02-06 17:48:21,994 INFO [optim.py:369] (0/4) 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 17:48:31,793 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=122479.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:48:36,701 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122486.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:48:43,463 INFO [train.py:901] (0/4) Epoch 16, batch 1250, loss[loss=0.213, simple_loss=0.2864, pruned_loss=0.06977, over 8459.00 frames. ], tot_loss[loss=0.2176, simple_loss=0.297, pruned_loss=0.06911, over 1620096.16 frames. ], batch size: 25, lr: 4.81e-03, grad_scale: 8.0 +2023-02-06 17:49:19,094 INFO [train.py:901] (0/4) Epoch 16, batch 1300, loss[loss=0.2187, simple_loss=0.2991, pruned_loss=0.06918, over 8297.00 frames. ], tot_loss[loss=0.2173, simple_loss=0.2975, pruned_loss=0.06858, over 1623337.50 frames. ], batch size: 23, lr: 4.81e-03, grad_scale: 8.0 +2023-02-06 17:49:26,978 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=122557.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:49:33,314 INFO [optim.py:369] (0/4) 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 17:49:52,150 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4991, 2.6126, 1.7409, 2.2194, 2.0983, 1.3609, 1.9797, 2.0571], + device='cuda:0'), covar=tensor([0.1642, 0.0437, 0.1303, 0.0664, 0.0766, 0.1598, 0.1126, 0.1070], + device='cuda:0'), in_proj_covar=tensor([0.0352, 0.0237, 0.0331, 0.0306, 0.0304, 0.0331, 0.0346, 0.0314], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 17:49:53,484 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4191, 2.3201, 4.5526, 1.9857, 2.6854, 5.1829, 5.2990, 4.4866], + device='cuda:0'), covar=tensor([0.1119, 0.1275, 0.0268, 0.1843, 0.1031, 0.0192, 0.0414, 0.0565], + device='cuda:0'), in_proj_covar=tensor([0.0281, 0.0309, 0.0273, 0.0302, 0.0295, 0.0249, 0.0381, 0.0299], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 17:49:54,923 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122594.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:49:56,067 INFO [train.py:901] (0/4) Epoch 16, batch 1350, loss[loss=0.2395, simple_loss=0.3212, pruned_loss=0.07888, over 8110.00 frames. ], tot_loss[loss=0.2175, simple_loss=0.2973, pruned_loss=0.06888, over 1616419.58 frames. ], batch size: 23, lr: 4.81e-03, grad_scale: 8.0 +2023-02-06 17:50:25,522 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9119, 2.1733, 1.7342, 2.5549, 1.2044, 1.5243, 1.7094, 2.0116], + device='cuda:0'), covar=tensor([0.0702, 0.0673, 0.0904, 0.0345, 0.1098, 0.1303, 0.0949, 0.0763], + device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0202, 0.0247, 0.0212, 0.0211, 0.0249, 0.0254, 0.0213], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 17:50:31,463 INFO [train.py:901] (0/4) Epoch 16, batch 1400, loss[loss=0.2138, simple_loss=0.3029, pruned_loss=0.06235, over 8339.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.2983, pruned_loss=0.06893, over 1621473.55 frames. ], batch size: 26, lr: 4.81e-03, grad_scale: 8.0 +2023-02-06 17:50:45,927 INFO [optim.py:369] (0/4) 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 17:50:49,404 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122672.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:50:55,471 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=122681.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:50:56,957 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122683.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:51:05,780 INFO [train.py:901] (0/4) Epoch 16, batch 1450, loss[loss=0.1809, simple_loss=0.2566, pruned_loss=0.05258, over 7700.00 frames. ], tot_loss[loss=0.2161, simple_loss=0.2958, pruned_loss=0.06814, over 1619977.53 frames. ], batch size: 18, lr: 4.81e-03, grad_scale: 4.0 +2023-02-06 17:51:12,710 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-06 17:51:15,649 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122708.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:51:26,183 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 +2023-02-06 17:51:26,747 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3557, 1.7425, 2.6968, 1.2301, 1.7289, 1.6943, 1.4889, 1.8174], + device='cuda:0'), covar=tensor([0.1898, 0.2360, 0.0878, 0.4510, 0.1925, 0.3254, 0.2147, 0.2227], + device='cuda:0'), in_proj_covar=tensor([0.0506, 0.0558, 0.0541, 0.0610, 0.0628, 0.0573, 0.0502, 0.0621], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 17:51:42,664 INFO [train.py:901] (0/4) Epoch 16, batch 1500, loss[loss=0.2434, simple_loss=0.3308, pruned_loss=0.07803, over 8354.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2958, pruned_loss=0.06856, over 1614902.37 frames. ], batch size: 24, lr: 4.81e-03, grad_scale: 4.0 +2023-02-06 17:51:56,875 INFO [optim.py:369] (0/4) 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 17:52:16,422 INFO [train.py:901] (0/4) Epoch 16, batch 1550, loss[loss=0.2304, simple_loss=0.2818, pruned_loss=0.08949, over 7696.00 frames. ], tot_loss[loss=0.2185, simple_loss=0.2975, pruned_loss=0.06974, over 1617399.40 frames. ], batch size: 18, lr: 4.81e-03, grad_scale: 4.0 +2023-02-06 17:52:16,601 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122796.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:52:41,663 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=122830.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:52:52,397 INFO [train.py:901] (0/4) Epoch 16, batch 1600, loss[loss=0.1849, simple_loss=0.2613, pruned_loss=0.05424, over 7783.00 frames. ], tot_loss[loss=0.2161, simple_loss=0.2948, pruned_loss=0.0687, over 1607961.60 frames. ], batch size: 19, lr: 4.81e-03, grad_scale: 8.0 +2023-02-06 17:52:55,408 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122850.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:53:07,648 INFO [optim.py:369] (0/4) 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] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122875.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:53:23,663 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122890.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:53:27,666 INFO [train.py:901] (0/4) Epoch 16, batch 1650, loss[loss=0.2148, simple_loss=0.2934, pruned_loss=0.06809, over 8089.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.2938, pruned_loss=0.06767, over 1608439.14 frames. ], batch size: 21, lr: 4.81e-03, grad_scale: 8.0 +2023-02-06 17:53:49,584 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122928.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:54:02,293 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.8056, 3.8015, 3.4616, 1.7721, 3.3771, 3.5087, 3.4097, 3.1982], + device='cuda:0'), covar=tensor([0.0970, 0.0697, 0.1111, 0.4809, 0.0892, 0.1089, 0.1451, 0.1020], + device='cuda:0'), in_proj_covar=tensor([0.0482, 0.0402, 0.0402, 0.0500, 0.0397, 0.0400, 0.0390, 0.0351], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 17:54:02,358 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122945.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:54:02,860 INFO [train.py:901] (0/4) Epoch 16, batch 1700, loss[loss=0.2671, simple_loss=0.34, pruned_loss=0.09711, over 8108.00 frames. ], tot_loss[loss=0.2153, simple_loss=0.2945, pruned_loss=0.06802, over 1609618.32 frames. ], batch size: 23, lr: 4.80e-03, grad_scale: 8.0 +2023-02-06 17:54:08,408 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122953.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:54:17,616 INFO [optim.py:369] (0/4) 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 17:54:38,073 INFO [train.py:901] (0/4) Epoch 16, batch 1750, loss[loss=0.2121, simple_loss=0.3006, pruned_loss=0.06179, over 8100.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.2953, pruned_loss=0.0679, over 1614984.53 frames. ], batch size: 23, lr: 4.80e-03, grad_scale: 8.0 +2023-02-06 17:55:00,433 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7222, 1.6794, 2.0813, 1.5260, 1.2503, 2.0950, 0.2907, 1.3203], + device='cuda:0'), covar=tensor([0.2267, 0.1775, 0.0451, 0.1626, 0.3237, 0.0430, 0.2615, 0.1565], + device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0180, 0.0111, 0.0212, 0.0255, 0.0115, 0.0161, 0.0176], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 17:55:12,109 INFO [train.py:901] (0/4) Epoch 16, batch 1800, loss[loss=0.2145, simple_loss=0.2944, pruned_loss=0.06727, over 8587.00 frames. ], tot_loss[loss=0.2159, simple_loss=0.2951, pruned_loss=0.06836, over 1610719.95 frames. ], batch size: 34, lr: 4.80e-03, grad_scale: 8.0 +2023-02-06 17:55:16,364 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=123052.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:55:18,540 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.19 vs. limit=5.0 +2023-02-06 17:55:27,705 INFO [optim.py:369] (0/4) 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 17:55:35,400 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=123077.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:55:48,779 INFO [train.py:901] (0/4) Epoch 16, batch 1850, loss[loss=0.2055, simple_loss=0.2898, pruned_loss=0.06056, over 7977.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2951, pruned_loss=0.06823, over 1615141.31 frames. ], batch size: 21, lr: 4.80e-03, grad_scale: 8.0 +2023-02-06 17:55:55,607 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0506, 1.6658, 2.0990, 1.8603, 1.9830, 2.0176, 1.8470, 0.7743], + device='cuda:0'), covar=tensor([0.4702, 0.4131, 0.1615, 0.2739, 0.2059, 0.2563, 0.1705, 0.4246], + device='cuda:0'), in_proj_covar=tensor([0.0911, 0.0922, 0.0758, 0.0894, 0.0956, 0.0843, 0.0720, 0.0797], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 17:55:58,183 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.8838, 6.1243, 5.2363, 2.5857, 5.3262, 5.7751, 5.5630, 5.3143], + device='cuda:0'), covar=tensor([0.0513, 0.0336, 0.0906, 0.3939, 0.0587, 0.0829, 0.1113, 0.0626], + device='cuda:0'), in_proj_covar=tensor([0.0495, 0.0411, 0.0413, 0.0513, 0.0408, 0.0411, 0.0401, 0.0360], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 17:56:22,235 INFO [train.py:901] (0/4) Epoch 16, batch 1900, loss[loss=0.2165, simple_loss=0.2883, pruned_loss=0.07233, over 7520.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.2948, pruned_loss=0.06819, over 1613622.91 frames. ], batch size: 18, lr: 4.80e-03, grad_scale: 8.0 +2023-02-06 17:56:31,862 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3770, 2.0005, 2.7422, 2.2350, 2.6170, 2.2667, 1.9819, 1.3750], + device='cuda:0'), covar=tensor([0.4519, 0.4218, 0.1537, 0.3153, 0.2117, 0.2644, 0.1820, 0.4713], + device='cuda:0'), in_proj_covar=tensor([0.0914, 0.0924, 0.0760, 0.0896, 0.0958, 0.0844, 0.0721, 0.0799], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 17:56:36,271 INFO [optim.py:369] (0/4) 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 17:56:50,196 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 +2023-02-06 17:56:57,741 INFO [train.py:901] (0/4) Epoch 16, batch 1950, loss[loss=0.2061, simple_loss=0.2816, pruned_loss=0.06527, over 7660.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.2955, pruned_loss=0.0686, over 1607072.50 frames. ], batch size: 19, lr: 4.80e-03, grad_scale: 8.0 +2023-02-06 17:56:59,131 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-06 17:57:01,383 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=123201.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:57:11,324 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-06 17:57:18,788 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=123226.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:57:24,071 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=123234.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 17:57:30,725 WARNING [train.py:1067] (0/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-06 17:57:32,091 INFO [train.py:901] (0/4) Epoch 16, batch 2000, loss[loss=0.206, simple_loss=0.2912, pruned_loss=0.0604, over 8189.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.294, pruned_loss=0.06762, over 1609569.11 frames. ], batch size: 23, lr: 4.80e-03, grad_scale: 8.0 +2023-02-06 17:57:46,346 INFO [optim.py:369] (0/4) 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 17:58:05,887 INFO [train.py:901] (0/4) Epoch 16, batch 2050, loss[loss=0.2131, simple_loss=0.2971, pruned_loss=0.06458, over 8182.00 frames. ], tot_loss[loss=0.215, simple_loss=0.2939, pruned_loss=0.06803, over 1601390.12 frames. ], batch size: 23, lr: 4.80e-03, grad_scale: 8.0 +2023-02-06 17:58:31,864 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 16, batch 2100, loss[loss=0.2335, simple_loss=0.2819, pruned_loss=0.0925, over 7533.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.2949, pruned_loss=0.06872, over 1605976.41 frames. ], batch size: 18, lr: 4.80e-03, grad_scale: 8.0 +2023-02-06 17:58:43,355 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 17:59:14,279 INFO [train.py:901] (0/4) Epoch 16, batch 2150, loss[loss=0.1734, simple_loss=0.2565, pruned_loss=0.04519, over 7267.00 frames. ], tot_loss[loss=0.2154, simple_loss=0.2945, pruned_loss=0.06817, over 1607721.28 frames. ], batch size: 16, lr: 4.80e-03, grad_scale: 8.0 +2023-02-06 17:59:19,805 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8805, 1.5845, 2.0567, 1.7832, 1.9776, 1.9248, 1.6680, 0.6857], + device='cuda:0'), covar=tensor([0.5444, 0.4572, 0.1671, 0.2929, 0.2225, 0.2785, 0.2034, 0.4956], + device='cuda:0'), in_proj_covar=tensor([0.0920, 0.0928, 0.0763, 0.0900, 0.0964, 0.0846, 0.0720, 0.0803], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 17:59:22,896 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.9293, 6.1457, 5.2203, 2.6212, 5.3154, 5.7554, 5.6532, 5.4231], + device='cuda:0'), covar=tensor([0.0622, 0.0336, 0.0940, 0.4299, 0.0740, 0.0692, 0.0922, 0.0446], + device='cuda:0'), in_proj_covar=tensor([0.0493, 0.0410, 0.0410, 0.0511, 0.0404, 0.0410, 0.0397, 0.0358], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 17:59:50,132 INFO [train.py:901] (0/4) Epoch 16, batch 2200, loss[loss=0.2149, simple_loss=0.2775, pruned_loss=0.07614, over 7439.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2947, pruned_loss=0.06844, over 1604024.17 frames. ], batch size: 17, lr: 4.79e-03, grad_scale: 8.0 +2023-02-06 17:59:50,270 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=123446.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:00:04,143 INFO [optim.py:369] (0/4) 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 18:00:23,384 INFO [train.py:901] (0/4) Epoch 16, batch 2250, loss[loss=0.2397, simple_loss=0.3326, pruned_loss=0.07338, over 8194.00 frames. ], tot_loss[loss=0.215, simple_loss=0.2944, pruned_loss=0.06782, over 1605766.41 frames. ], batch size: 23, lr: 4.79e-03, grad_scale: 8.0 +2023-02-06 18:00:46,379 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 +2023-02-06 18:00:58,617 INFO [train.py:901] (0/4) Epoch 16, batch 2300, loss[loss=0.2602, simple_loss=0.3388, pruned_loss=0.09082, over 8497.00 frames. ], tot_loss[loss=0.216, simple_loss=0.2954, pruned_loss=0.06828, over 1611338.50 frames. ], batch size: 26, lr: 4.79e-03, grad_scale: 8.0 +2023-02-06 18:01:13,217 INFO [optim.py:369] (0/4) 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 18:01:22,764 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6009, 1.9764, 2.0391, 1.2891, 2.2087, 1.5658, 0.5521, 1.8645], + device='cuda:0'), covar=tensor([0.0515, 0.0256, 0.0207, 0.0474, 0.0257, 0.0707, 0.0727, 0.0237], + device='cuda:0'), in_proj_covar=tensor([0.0423, 0.0364, 0.0313, 0.0416, 0.0350, 0.0507, 0.0372, 0.0389], + device='cuda:0'), 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:0') +2023-02-06 18:01:32,631 INFO [train.py:901] (0/4) Epoch 16, batch 2350, loss[loss=0.2054, simple_loss=0.2752, pruned_loss=0.06783, over 7699.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.2953, pruned_loss=0.06862, over 1609304.43 frames. ], batch size: 18, lr: 4.79e-03, grad_scale: 8.0 +2023-02-06 18:01:38,854 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=123605.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:01:55,675 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=123630.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:02:06,132 INFO [train.py:901] (0/4) Epoch 16, batch 2400, loss[loss=0.2158, simple_loss=0.2863, pruned_loss=0.07265, over 7992.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.2957, pruned_loss=0.06894, over 1611082.33 frames. ], batch size: 21, lr: 4.79e-03, grad_scale: 8.0 +2023-02-06 18:02:22,334 INFO [optim.py:369] (0/4) 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 18:02:28,428 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=123676.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:02:42,475 INFO [train.py:901] (0/4) Epoch 16, batch 2450, loss[loss=0.233, simple_loss=0.3131, pruned_loss=0.0764, over 8452.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.2972, pruned_loss=0.06969, over 1614467.83 frames. ], batch size: 27, lr: 4.79e-03, grad_scale: 8.0 +2023-02-06 18:03:15,344 INFO [train.py:901] (0/4) Epoch 16, batch 2500, loss[loss=0.2366, simple_loss=0.3302, pruned_loss=0.07155, over 8256.00 frames. ], tot_loss[loss=0.2196, simple_loss=0.2987, pruned_loss=0.07027, over 1619243.86 frames. ], batch size: 22, lr: 4.79e-03, grad_scale: 8.0 +2023-02-06 18:03:25,452 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0468, 1.3789, 1.6240, 1.2372, 1.0393, 1.4104, 1.6589, 1.4953], + device='cuda:0'), covar=tensor([0.0523, 0.1325, 0.1694, 0.1508, 0.0594, 0.1575, 0.0700, 0.0651], + device='cuda:0'), in_proj_covar=tensor([0.0098, 0.0152, 0.0190, 0.0157, 0.0100, 0.0163, 0.0114, 0.0140], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 18:03:29,366 INFO [optim.py:369] (0/4) 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 18:03:46,943 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=123790.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:03:47,734 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 16, batch 2550, loss[loss=0.2628, simple_loss=0.3352, pruned_loss=0.09524, over 7406.00 frames. ], tot_loss[loss=0.2182, simple_loss=0.2975, pruned_loss=0.06939, over 1616223.34 frames. ], batch size: 71, lr: 4.79e-03, grad_scale: 8.0 +2023-02-06 18:04:24,936 INFO [train.py:901] (0/4) Epoch 16, batch 2600, loss[loss=0.2576, simple_loss=0.3248, pruned_loss=0.09516, over 7305.00 frames. ], tot_loss[loss=0.2171, simple_loss=0.2967, pruned_loss=0.06873, over 1615978.15 frames. ], batch size: 72, lr: 4.79e-03, grad_scale: 8.0 +2023-02-06 18:04:38,916 INFO [optim.py:369] (0/4) 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 18:04:54,356 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 16, batch 2650, loss[loss=0.2249, simple_loss=0.3056, pruned_loss=0.07207, over 8035.00 frames. ], tot_loss[loss=0.2185, simple_loss=0.298, pruned_loss=0.0695, over 1618479.26 frames. ], batch size: 20, lr: 4.79e-03, grad_scale: 8.0 +2023-02-06 18:05:06,333 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 16, batch 2700, loss[loss=0.1991, simple_loss=0.2691, pruned_loss=0.06453, over 7693.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.2975, pruned_loss=0.06939, over 1618442.68 frames. ], batch size: 18, lr: 4.79e-03, grad_scale: 8.0 +2023-02-06 18:05:48,215 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 16, batch 2750, loss[loss=0.1751, simple_loss=0.2555, pruned_loss=0.04737, over 7644.00 frames. ], tot_loss[loss=0.2195, simple_loss=0.2984, pruned_loss=0.07029, over 1613852.24 frames. ], batch size: 19, lr: 4.78e-03, grad_scale: 8.0 +2023-02-06 18:06:10,309 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-124000.pt +2023-02-06 18:06:45,090 INFO [train.py:901] (0/4) Epoch 16, batch 2800, loss[loss=0.1926, simple_loss=0.2743, pruned_loss=0.05539, over 8089.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.297, pruned_loss=0.06932, over 1613810.05 frames. ], batch size: 21, lr: 4.78e-03, grad_scale: 8.0 +2023-02-06 18:06:46,000 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124047.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:06:50,830 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124054.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:06:59,432 INFO [optim.py:369] (0/4) 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 18:07:03,050 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124072.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:07:08,249 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124080.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:07:18,991 INFO [train.py:901] (0/4) Epoch 16, batch 2850, loss[loss=0.2054, simple_loss=0.2866, pruned_loss=0.06206, over 8470.00 frames. ], tot_loss[loss=0.2179, simple_loss=0.2973, pruned_loss=0.0692, over 1614911.20 frames. ], batch size: 27, lr: 4.78e-03, grad_scale: 8.0 +2023-02-06 18:07:22,869 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 +2023-02-06 18:07:55,341 INFO [train.py:901] (0/4) Epoch 16, batch 2900, loss[loss=0.1813, simple_loss=0.2597, pruned_loss=0.05142, over 7795.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.2971, pruned_loss=0.06952, over 1610070.53 frames. ], batch size: 20, lr: 4.78e-03, grad_scale: 8.0 +2023-02-06 18:08:06,256 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124161.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:08:06,292 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2135, 1.9363, 2.6008, 2.1845, 2.4227, 2.2146, 1.8777, 1.2898], + device='cuda:0'), covar=tensor([0.4765, 0.4418, 0.1511, 0.3013, 0.2223, 0.2598, 0.1822, 0.4596], + device='cuda:0'), in_proj_covar=tensor([0.0914, 0.0924, 0.0758, 0.0894, 0.0961, 0.0848, 0.0723, 0.0797], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 18:08:10,031 INFO [optim.py:369] (0/4) 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 18:08:22,880 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 16, batch 2950, loss[loss=0.1816, simple_loss=0.2616, pruned_loss=0.05078, over 7793.00 frames. ], tot_loss[loss=0.2192, simple_loss=0.2979, pruned_loss=0.07022, over 1608602.76 frames. ], batch size: 19, lr: 4.78e-03, grad_scale: 8.0 +2023-02-06 18:08:29,750 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2915, 2.7108, 3.1464, 1.6442, 3.3504, 2.0182, 1.5400, 2.2152], + device='cuda:0'), covar=tensor([0.0617, 0.0295, 0.0195, 0.0659, 0.0273, 0.0664, 0.0737, 0.0431], + device='cuda:0'), in_proj_covar=tensor([0.0425, 0.0363, 0.0311, 0.0417, 0.0350, 0.0508, 0.0371, 0.0387], + device='cuda:0'), 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:0') +2023-02-06 18:08:35,635 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-06 18:08:55,340 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=124234.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:09:01,459 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.77 vs. limit=5.0 +2023-02-06 18:09:03,840 INFO [train.py:901] (0/4) Epoch 16, batch 3000, loss[loss=0.2457, simple_loss=0.322, pruned_loss=0.0847, over 8355.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.2979, pruned_loss=0.07009, over 1611640.21 frames. ], batch size: 24, lr: 4.78e-03, grad_scale: 8.0 +2023-02-06 18:09:03,841 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 18:09:16,270 INFO [train.py:935] (0/4) Epoch 16, validation: loss=0.1794, simple_loss=0.2796, pruned_loss=0.03958, over 944034.00 frames. +2023-02-06 18:09:16,271 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-06 18:09:32,708 INFO [optim.py:369] (0/4) 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 18:09:49,092 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124290.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:09:52,918 INFO [train.py:901] (0/4) Epoch 16, batch 3050, loss[loss=0.2094, simple_loss=0.2971, pruned_loss=0.0609, over 8453.00 frames. ], tot_loss[loss=0.2202, simple_loss=0.2993, pruned_loss=0.07056, over 1615752.21 frames. ], batch size: 29, lr: 4.78e-03, grad_scale: 8.0 +2023-02-06 18:10:16,794 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.9619, 1.9697, 6.0581, 2.1141, 5.4775, 5.0460, 5.5987, 5.4659], + device='cuda:0'), covar=tensor([0.0440, 0.4413, 0.0288, 0.3922, 0.0824, 0.0829, 0.0463, 0.0450], + device='cuda:0'), in_proj_covar=tensor([0.0556, 0.0607, 0.0637, 0.0588, 0.0659, 0.0568, 0.0561, 0.0623], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 18:10:25,965 INFO [train.py:901] (0/4) Epoch 16, batch 3100, loss[loss=0.2509, simple_loss=0.3228, pruned_loss=0.08947, over 8315.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.299, pruned_loss=0.06987, over 1617696.95 frames. ], batch size: 25, lr: 4.78e-03, grad_scale: 8.0 +2023-02-06 18:10:28,058 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124349.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:10:39,304 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124366.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 18:10:39,810 INFO [optim.py:369] (0/4) 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 18:11:01,482 INFO [train.py:901] (0/4) Epoch 16, batch 3150, loss[loss=0.2572, simple_loss=0.3391, pruned_loss=0.08767, over 8107.00 frames. ], tot_loss[loss=0.2196, simple_loss=0.2993, pruned_loss=0.06996, over 1616130.71 frames. ], batch size: 23, lr: 4.78e-03, grad_scale: 8.0 +2023-02-06 18:11:02,949 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=124398.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:11:21,457 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=124424.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:11:23,925 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.40 vs. limit=5.0 +2023-02-06 18:11:36,607 INFO [train.py:901] (0/4) Epoch 16, batch 3200, loss[loss=0.1731, simple_loss=0.2582, pruned_loss=0.04401, over 7806.00 frames. ], tot_loss[loss=0.2201, simple_loss=0.2997, pruned_loss=0.07027, over 1616677.16 frames. ], batch size: 19, lr: 4.78e-03, grad_scale: 8.0 +2023-02-06 18:11:39,392 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124450.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:11:50,434 INFO [optim.py:369] (0/4) 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 18:12:09,965 INFO [train.py:901] (0/4) Epoch 16, batch 3250, loss[loss=0.2342, simple_loss=0.3211, pruned_loss=0.07359, over 8188.00 frames. ], tot_loss[loss=0.2197, simple_loss=0.2995, pruned_loss=0.06991, over 1620992.03 frames. ], batch size: 23, lr: 4.77e-03, grad_scale: 8.0 +2023-02-06 18:12:23,011 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124513.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:12:38,295 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4180, 2.0970, 2.9029, 2.3577, 2.8357, 2.2891, 2.0578, 1.7760], + device='cuda:0'), covar=tensor([0.4448, 0.4455, 0.1509, 0.3135, 0.2109, 0.2544, 0.1707, 0.4545], + device='cuda:0'), in_proj_covar=tensor([0.0913, 0.0922, 0.0756, 0.0893, 0.0960, 0.0846, 0.0721, 0.0796], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 18:12:40,898 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124539.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:12:42,554 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 +2023-02-06 18:12:45,394 INFO [train.py:901] (0/4) Epoch 16, batch 3300, loss[loss=0.2716, simple_loss=0.3387, pruned_loss=0.1022, over 6779.00 frames. ], tot_loss[loss=0.2186, simple_loss=0.298, pruned_loss=0.06953, over 1611818.89 frames. ], batch size: 72, lr: 4.77e-03, grad_scale: 8.0 +2023-02-06 18:12:59,411 INFO [optim.py:369] (0/4) 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 18:13:18,831 INFO [train.py:901] (0/4) Epoch 16, batch 3350, loss[loss=0.2606, simple_loss=0.3251, pruned_loss=0.09806, over 6889.00 frames. ], tot_loss[loss=0.2188, simple_loss=0.2982, pruned_loss=0.06969, over 1610525.32 frames. ], batch size: 73, lr: 4.77e-03, grad_scale: 8.0 +2023-02-06 18:13:25,335 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124605.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:13:43,462 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124630.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:13:46,005 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=124634.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:13:54,702 INFO [train.py:901] (0/4) Epoch 16, batch 3400, loss[loss=0.211, simple_loss=0.3018, pruned_loss=0.06014, over 8474.00 frames. ], tot_loss[loss=0.2185, simple_loss=0.2979, pruned_loss=0.06955, over 1613058.94 frames. ], batch size: 25, lr: 4.77e-03, grad_scale: 8.0 +2023-02-06 18:14:01,589 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124656.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:14:08,831 INFO [optim.py:369] (0/4) 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 18:14:18,569 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124681.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:14:28,870 INFO [train.py:901] (0/4) Epoch 16, batch 3450, loss[loss=0.2143, simple_loss=0.2916, pruned_loss=0.06849, over 8550.00 frames. ], tot_loss[loss=0.2187, simple_loss=0.2983, pruned_loss=0.06958, over 1615303.40 frames. ], batch size: 49, lr: 4.77e-03, grad_scale: 16.0 +2023-02-06 18:14:38,557 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=124710.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 18:14:47,104 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.07 vs. limit=5.0 +2023-02-06 18:15:05,352 INFO [train.py:901] (0/4) Epoch 16, batch 3500, loss[loss=0.2006, simple_loss=0.2839, pruned_loss=0.05864, over 8134.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.297, pruned_loss=0.06889, over 1613707.05 frames. ], batch size: 22, lr: 4.77e-03, grad_scale: 16.0 +2023-02-06 18:15:07,649 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124749.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:15:18,017 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3161, 2.3399, 1.5565, 1.9765, 1.9048, 1.2549, 1.6383, 1.8497], + device='cuda:0'), covar=tensor([0.1570, 0.0457, 0.1408, 0.0688, 0.0804, 0.1966, 0.1288, 0.0960], + device='cuda:0'), in_proj_covar=tensor([0.0353, 0.0234, 0.0327, 0.0305, 0.0300, 0.0334, 0.0348, 0.0316], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 18:15:20,550 INFO [optim.py:369] (0/4) 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 18:15:20,977 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 +2023-02-06 18:15:22,091 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124769.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:15:38,226 WARNING [train.py:1067] (0/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-06 18:15:38,975 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=124794.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:15:39,088 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124794.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:15:39,723 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124795.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:15:40,177 INFO [train.py:901] (0/4) Epoch 16, batch 3550, loss[loss=0.2126, simple_loss=0.2938, pruned_loss=0.06571, over 8496.00 frames. ], tot_loss[loss=0.2184, simple_loss=0.2981, pruned_loss=0.06941, over 1619907.39 frames. ], batch size: 28, lr: 4.77e-03, grad_scale: 16.0 +2023-02-06 18:15:56,871 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124820.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:16:00,135 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124825.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 18:16:08,884 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124838.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:16:14,221 INFO [train.py:901] (0/4) Epoch 16, batch 3600, loss[loss=0.1904, simple_loss=0.2653, pruned_loss=0.05777, over 7702.00 frames. ], tot_loss[loss=0.2187, simple_loss=0.2983, pruned_loss=0.06956, over 1617352.37 frames. ], batch size: 18, lr: 4.77e-03, grad_scale: 16.0 +2023-02-06 18:16:26,986 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.08 vs. limit=5.0 +2023-02-06 18:16:30,804 INFO [optim.py:369] (0/4) 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 18:16:50,929 INFO [train.py:901] (0/4) Epoch 16, batch 3650, loss[loss=0.2517, simple_loss=0.3291, pruned_loss=0.08713, over 8360.00 frames. ], tot_loss[loss=0.2199, simple_loss=0.2994, pruned_loss=0.07022, over 1622580.31 frames. ], batch size: 24, lr: 4.77e-03, grad_scale: 16.0 +2023-02-06 18:16:53,007 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.3665, 1.3101, 4.5456, 1.7739, 4.0133, 3.8117, 4.0720, 3.9913], + device='cuda:0'), covar=tensor([0.0517, 0.4708, 0.0442, 0.3893, 0.1061, 0.0906, 0.0582, 0.0632], + device='cuda:0'), in_proj_covar=tensor([0.0557, 0.0613, 0.0635, 0.0587, 0.0660, 0.0566, 0.0561, 0.0623], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 18:16:59,886 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 16, batch 3700, loss[loss=0.22, simple_loss=0.2933, pruned_loss=0.07336, over 8088.00 frames. ], tot_loss[loss=0.2189, simple_loss=0.2986, pruned_loss=0.06962, over 1623698.87 frames. ], batch size: 21, lr: 4.77e-03, grad_scale: 16.0 +2023-02-06 18:17:38,859 WARNING [train.py:1067] (0/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-06 18:17:40,141 INFO [optim.py:369] (0/4) 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 18:17:47,827 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4605, 1.8707, 3.0494, 1.3437, 2.2882, 1.8641, 1.6060, 2.1931], + device='cuda:0'), covar=tensor([0.1968, 0.2317, 0.0685, 0.4301, 0.1589, 0.2992, 0.2145, 0.2128], + device='cuda:0'), in_proj_covar=tensor([0.0505, 0.0559, 0.0541, 0.0611, 0.0630, 0.0569, 0.0500, 0.0618], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 18:18:01,619 INFO [train.py:901] (0/4) Epoch 16, batch 3750, loss[loss=0.2097, simple_loss=0.2861, pruned_loss=0.06659, over 8089.00 frames. ], tot_loss[loss=0.2171, simple_loss=0.2971, pruned_loss=0.06857, over 1618862.90 frames. ], batch size: 21, lr: 4.77e-03, grad_scale: 16.0 +2023-02-06 18:18:04,452 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=125000.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:18:07,873 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125005.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:18:20,983 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=125025.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:18:22,459 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.6334, 1.7201, 4.8423, 1.9791, 4.2949, 4.0731, 4.3711, 4.2271], + device='cuda:0'), covar=tensor([0.0537, 0.4111, 0.0436, 0.3436, 0.0983, 0.0855, 0.0511, 0.0611], + device='cuda:0'), in_proj_covar=tensor([0.0556, 0.0612, 0.0635, 0.0584, 0.0658, 0.0565, 0.0561, 0.0624], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 18:18:24,558 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125030.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:18:35,176 INFO [train.py:901] (0/4) Epoch 16, batch 3800, loss[loss=0.17, simple_loss=0.2502, pruned_loss=0.04491, over 7684.00 frames. ], tot_loss[loss=0.2171, simple_loss=0.2968, pruned_loss=0.0687, over 1619349.10 frames. ], batch size: 18, lr: 4.76e-03, grad_scale: 16.0 +2023-02-06 18:18:49,284 INFO [optim.py:369] (0/4) 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] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125081.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 18:19:10,727 INFO [train.py:901] (0/4) Epoch 16, batch 3850, loss[loss=0.2297, simple_loss=0.3136, pruned_loss=0.07292, over 8568.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.2973, pruned_loss=0.06916, over 1621389.55 frames. ], batch size: 31, lr: 4.76e-03, grad_scale: 16.0 +2023-02-06 18:19:18,419 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125106.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 18:19:19,270 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 +2023-02-06 18:19:24,389 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125115.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:19:41,019 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125140.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:19:45,073 INFO [train.py:901] (0/4) Epoch 16, batch 3900, loss[loss=0.2019, simple_loss=0.2916, pruned_loss=0.05608, over 8098.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.2961, pruned_loss=0.06871, over 1618439.75 frames. ], batch size: 21, lr: 4.76e-03, grad_scale: 16.0 +2023-02-06 18:19:45,091 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-06 18:19:45,214 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125146.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:19:52,241 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 +2023-02-06 18:19:58,178 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125165.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:19:59,303 INFO [optim.py:369] (0/4) 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 18:20:09,606 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=125182.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:20:15,220 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 16, batch 3950, loss[loss=0.2056, simple_loss=0.2844, pruned_loss=0.0634, over 8087.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2963, pruned_loss=0.06856, over 1619331.69 frames. ], batch size: 21, lr: 4.76e-03, grad_scale: 16.0 +2023-02-06 18:20:20,238 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 +2023-02-06 18:20:55,581 INFO [train.py:901] (0/4) Epoch 16, batch 4000, loss[loss=0.2014, simple_loss=0.2869, pruned_loss=0.05795, over 7971.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.2963, pruned_loss=0.06864, over 1616100.04 frames. ], batch size: 21, lr: 4.76e-03, grad_scale: 16.0 +2023-02-06 18:21:09,904 INFO [optim.py:369] (0/4) 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 18:21:10,312 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.73 vs. limit=5.0 +2023-02-06 18:21:29,135 INFO [train.py:901] (0/4) Epoch 16, batch 4050, loss[loss=0.2237, simple_loss=0.3088, pruned_loss=0.06926, over 8327.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.2978, pruned_loss=0.06936, over 1619121.77 frames. ], batch size: 25, lr: 4.76e-03, grad_scale: 16.0 +2023-02-06 18:21:29,952 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125297.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:21:43,914 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3623, 1.6395, 1.7420, 1.0495, 1.7950, 1.3419, 0.2149, 1.5591], + device='cuda:0'), covar=tensor([0.0363, 0.0253, 0.0186, 0.0340, 0.0302, 0.0650, 0.0602, 0.0187], + device='cuda:0'), in_proj_covar=tensor([0.0424, 0.0360, 0.0310, 0.0417, 0.0351, 0.0507, 0.0370, 0.0389], + device='cuda:0'), out_proj_covar=tensor([1.1663e-04, 9.6111e-05, 8.2316e-05, 1.1197e-04, 9.4792e-05, 1.4667e-04, + 1.0140e-04, 1.0537e-04], device='cuda:0') +2023-02-06 18:22:05,146 INFO [train.py:901] (0/4) Epoch 16, batch 4100, loss[loss=0.1828, simple_loss=0.2637, pruned_loss=0.051, over 7647.00 frames. ], tot_loss[loss=0.2185, simple_loss=0.2975, pruned_loss=0.06978, over 1616568.65 frames. ], batch size: 19, lr: 4.76e-03, grad_scale: 16.0 +2023-02-06 18:22:19,362 INFO [optim.py:369] (0/4) 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 18:22:20,410 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 +2023-02-06 18:22:22,350 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125371.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:22:24,205 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125374.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:22:37,712 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125394.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:22:38,920 INFO [train.py:901] (0/4) Epoch 16, batch 4150, loss[loss=0.2644, simple_loss=0.3431, pruned_loss=0.09282, over 8367.00 frames. ], tot_loss[loss=0.2192, simple_loss=0.2978, pruned_loss=0.07028, over 1609957.10 frames. ], batch size: 50, lr: 4.76e-03, grad_scale: 16.0 +2023-02-06 18:22:39,119 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125396.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:22:39,137 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125396.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:22:55,747 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125421.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:22:59,769 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.6789, 4.6945, 4.2087, 1.9525, 4.2039, 4.2314, 4.2590, 3.9932], + device='cuda:0'), covar=tensor([0.0752, 0.0497, 0.1067, 0.4714, 0.0901, 0.0928, 0.1321, 0.0791], + device='cuda:0'), in_proj_covar=tensor([0.0495, 0.0412, 0.0412, 0.0514, 0.0405, 0.0411, 0.0401, 0.0358], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 18:23:02,563 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4945, 1.4743, 1.8082, 1.2706, 1.1079, 1.7553, 0.1088, 1.1513], + device='cuda:0'), covar=tensor([0.2178, 0.1580, 0.0429, 0.1320, 0.3268, 0.0556, 0.2655, 0.1512], + device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0179, 0.0113, 0.0212, 0.0255, 0.0116, 0.0162, 0.0175], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 18:23:09,066 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125439.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:23:14,198 INFO [train.py:901] (0/4) Epoch 16, batch 4200, loss[loss=0.237, simple_loss=0.3154, pruned_loss=0.07927, over 8489.00 frames. ], tot_loss[loss=0.2195, simple_loss=0.2983, pruned_loss=0.07038, over 1609042.83 frames. ], batch size: 28, lr: 4.76e-03, grad_scale: 16.0 +2023-02-06 18:23:29,128 INFO [optim.py:369] (0/4) 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 18:23:39,902 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-06 18:23:44,590 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 16, batch 4250, loss[loss=0.1976, simple_loss=0.2761, pruned_loss=0.05956, over 7975.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.2972, pruned_loss=0.0695, over 1606983.08 frames. ], batch size: 21, lr: 4.76e-03, grad_scale: 16.0 +2023-02-06 18:24:01,601 WARNING [train.py:1067] (0/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-06 18:24:23,119 INFO [train.py:901] (0/4) Epoch 16, batch 4300, loss[loss=0.2504, simple_loss=0.3292, pruned_loss=0.08577, over 8107.00 frames. ], tot_loss[loss=0.2189, simple_loss=0.298, pruned_loss=0.0699, over 1608309.90 frames. ], batch size: 23, lr: 4.75e-03, grad_scale: 16.0 +2023-02-06 18:24:24,746 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8621, 1.6090, 2.0350, 1.7817, 2.0401, 1.8847, 1.6508, 0.8357], + device='cuda:0'), covar=tensor([0.5107, 0.4173, 0.1644, 0.2952, 0.2009, 0.2592, 0.1764, 0.4306], + device='cuda:0'), in_proj_covar=tensor([0.0916, 0.0923, 0.0760, 0.0900, 0.0963, 0.0852, 0.0722, 0.0797], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 18:24:28,621 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125553.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:24:37,212 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6133, 1.3825, 2.3619, 1.1514, 2.1061, 2.4633, 2.6288, 2.1070], + device='cuda:0'), covar=tensor([0.0940, 0.1320, 0.0461, 0.2180, 0.0718, 0.0410, 0.0650, 0.0753], + device='cuda:0'), in_proj_covar=tensor([0.0278, 0.0309, 0.0274, 0.0299, 0.0290, 0.0249, 0.0383, 0.0296], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-06 18:24:38,404 INFO [optim.py:369] (0/4) 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 18:24:46,753 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125578.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:24:58,881 INFO [train.py:901] (0/4) Epoch 16, batch 4350, loss[loss=0.2388, simple_loss=0.2917, pruned_loss=0.0929, over 7648.00 frames. ], tot_loss[loss=0.2205, simple_loss=0.299, pruned_loss=0.07102, over 1610085.38 frames. ], batch size: 19, lr: 4.75e-03, grad_scale: 16.0 +2023-02-06 18:25:05,360 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125605.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:25:16,547 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 16, batch 4400, loss[loss=0.2117, simple_loss=0.283, pruned_loss=0.07019, over 7534.00 frames. ], tot_loss[loss=0.2209, simple_loss=0.2993, pruned_loss=0.0712, over 1611469.60 frames. ], batch size: 18, lr: 4.75e-03, grad_scale: 16.0 +2023-02-06 18:25:34,019 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-06 18:25:48,658 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 16, batch 4450, loss[loss=0.2111, simple_loss=0.3036, pruned_loss=0.0593, over 8313.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.2982, pruned_loss=0.06998, over 1613477.10 frames. ], batch size: 25, lr: 4.75e-03, grad_scale: 16.0 +2023-02-06 18:26:14,203 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-06 18:26:24,207 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=125718.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:26:38,221 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=125738.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:26:41,742 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125743.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:26:43,637 INFO [train.py:901] (0/4) Epoch 16, batch 4500, loss[loss=0.1985, simple_loss=0.273, pruned_loss=0.06199, over 7925.00 frames. ], tot_loss[loss=0.2189, simple_loss=0.2983, pruned_loss=0.06977, over 1616078.71 frames. ], batch size: 20, lr: 4.75e-03, grad_scale: 16.0 +2023-02-06 18:26:57,808 INFO [optim.py:369] (0/4) 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 18:27:04,075 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-06 18:27:10,712 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 16, batch 4550, loss[loss=0.2286, simple_loss=0.3128, pruned_loss=0.07222, over 8188.00 frames. ], tot_loss[loss=0.219, simple_loss=0.2984, pruned_loss=0.06985, over 1614056.20 frames. ], batch size: 23, lr: 4.75e-03, grad_scale: 16.0 +2023-02-06 18:27:45,720 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125833.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:27:54,474 INFO [train.py:901] (0/4) Epoch 16, batch 4600, loss[loss=0.2326, simple_loss=0.3188, pruned_loss=0.07319, over 8453.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.2986, pruned_loss=0.07005, over 1614679.47 frames. ], batch size: 48, lr: 4.75e-03, grad_scale: 16.0 +2023-02-06 18:27:59,476 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125853.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:28:05,144 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125861.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:28:08,979 INFO [optim.py:369] (0/4) 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 18:28:22,182 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125886.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:28:30,029 INFO [train.py:901] (0/4) Epoch 16, batch 4650, loss[loss=0.2217, simple_loss=0.2839, pruned_loss=0.0798, over 7648.00 frames. ], tot_loss[loss=0.2204, simple_loss=0.2994, pruned_loss=0.07072, over 1616241.78 frames. ], batch size: 19, lr: 4.75e-03, grad_scale: 16.0 +2023-02-06 18:28:31,607 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125898.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:28:54,787 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 +2023-02-06 18:29:06,119 INFO [train.py:901] (0/4) Epoch 16, batch 4700, loss[loss=0.1784, simple_loss=0.2652, pruned_loss=0.04582, over 8099.00 frames. ], tot_loss[loss=0.2209, simple_loss=0.2998, pruned_loss=0.07094, over 1617713.91 frames. ], batch size: 23, lr: 4.75e-03, grad_scale: 16.0 +2023-02-06 18:29:18,968 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=125965.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:29:20,237 INFO [optim.py:369] (0/4) 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 18:29:25,563 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.32 vs. limit=5.0 +2023-02-06 18:29:39,834 INFO [train.py:901] (0/4) Epoch 16, batch 4750, loss[loss=0.1906, simple_loss=0.2755, pruned_loss=0.05283, over 8078.00 frames. ], tot_loss[loss=0.2192, simple_loss=0.2987, pruned_loss=0.06984, over 1619017.43 frames. ], batch size: 21, lr: 4.75e-03, grad_scale: 16.0 +2023-02-06 18:29:42,623 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-126000.pt +2023-02-06 18:29:50,563 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4392, 2.5690, 2.0237, 2.2006, 2.0676, 1.6800, 2.1661, 2.0856], + device='cuda:0'), covar=tensor([0.1434, 0.0360, 0.0978, 0.0599, 0.0736, 0.1389, 0.0867, 0.0927], + device='cuda:0'), in_proj_covar=tensor([0.0349, 0.0230, 0.0322, 0.0301, 0.0297, 0.0328, 0.0340, 0.0311], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 18:29:55,988 INFO [zipformer.py:1185] (0/4) 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] (0/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-06 18:30:13,728 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-06 18:30:16,331 INFO [train.py:901] (0/4) Epoch 16, batch 4800, loss[loss=0.1901, simple_loss=0.2608, pruned_loss=0.05974, over 7210.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.2977, pruned_loss=0.06923, over 1618932.67 frames. ], batch size: 16, lr: 4.75e-03, grad_scale: 16.0 +2023-02-06 18:30:31,318 INFO [optim.py:369] (0/4) 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 18:30:40,360 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126080.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:30:44,993 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=126087.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:30:46,459 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 16, batch 4850, loss[loss=0.2217, simple_loss=0.3018, pruned_loss=0.07083, over 8462.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.2973, pruned_loss=0.06913, over 1616159.70 frames. ], batch size: 27, lr: 4.74e-03, grad_scale: 16.0 +2023-02-06 18:30:59,983 INFO [zipformer.py:1185] (0/4) 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] (0/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-06 18:31:03,310 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=126134.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:31:26,989 INFO [train.py:901] (0/4) Epoch 16, batch 4900, loss[loss=0.2114, simple_loss=0.2968, pruned_loss=0.06302, over 8335.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.2966, pruned_loss=0.0685, over 1615214.64 frames. ], batch size: 25, lr: 4.74e-03, grad_scale: 16.0 +2023-02-06 18:31:32,564 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126154.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:31:41,754 INFO [optim.py:369] (0/4) 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 18:31:50,103 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=126179.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:32:01,582 INFO [train.py:901] (0/4) Epoch 16, batch 4950, loss[loss=0.2538, simple_loss=0.3168, pruned_loss=0.09546, over 7000.00 frames. ], tot_loss[loss=0.2169, simple_loss=0.2966, pruned_loss=0.06864, over 1612540.29 frames. ], batch size: 74, lr: 4.74e-03, grad_scale: 16.0 +2023-02-06 18:32:06,028 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126202.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:32:35,839 INFO [train.py:901] (0/4) Epoch 16, batch 5000, loss[loss=0.1771, simple_loss=0.2592, pruned_loss=0.04746, over 8246.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.2953, pruned_loss=0.06786, over 1609292.04 frames. ], batch size: 22, lr: 4.74e-03, grad_scale: 16.0 +2023-02-06 18:32:50,294 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 16, batch 5050, loss[loss=0.1839, simple_loss=0.2605, pruned_loss=0.05368, over 7415.00 frames. ], tot_loss[loss=0.2153, simple_loss=0.2946, pruned_loss=0.068, over 1604801.05 frames. ], batch size: 17, lr: 4.74e-03, grad_scale: 8.0 +2023-02-06 18:33:38,217 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126336.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:33:38,802 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.4014, 4.3575, 3.9258, 1.9288, 3.8817, 3.9322, 3.9527, 3.6328], + device='cuda:0'), covar=tensor([0.0735, 0.0581, 0.1108, 0.4963, 0.0965, 0.1084, 0.1243, 0.1013], + device='cuda:0'), in_proj_covar=tensor([0.0496, 0.0412, 0.0413, 0.0514, 0.0404, 0.0409, 0.0402, 0.0357], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 18:33:41,491 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-06 18:33:44,851 INFO [train.py:901] (0/4) Epoch 16, batch 5100, loss[loss=0.217, simple_loss=0.3051, pruned_loss=0.06445, over 8140.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.295, pruned_loss=0.06822, over 1607591.98 frames. ], batch size: 22, lr: 4.74e-03, grad_scale: 8.0 +2023-02-06 18:33:55,151 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=126360.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:33:55,973 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=126361.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:34:01,122 INFO [optim.py:369] (0/4) 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 18:34:20,688 INFO [train.py:901] (0/4) Epoch 16, batch 5150, loss[loss=0.2152, simple_loss=0.2954, pruned_loss=0.06752, over 7780.00 frames. ], tot_loss[loss=0.2159, simple_loss=0.2953, pruned_loss=0.06822, over 1611596.00 frames. ], batch size: 19, lr: 4.74e-03, grad_scale: 8.0 +2023-02-06 18:34:22,775 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=126398.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:34:25,577 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.6945, 4.7069, 4.1810, 2.1014, 4.1744, 4.3298, 4.2694, 4.0401], + device='cuda:0'), covar=tensor([0.0691, 0.0564, 0.1131, 0.4476, 0.0884, 0.0829, 0.1234, 0.0808], + device='cuda:0'), in_proj_covar=tensor([0.0495, 0.0413, 0.0412, 0.0513, 0.0403, 0.0409, 0.0400, 0.0355], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 18:34:29,713 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.1941, 1.0823, 1.2699, 1.0568, 0.9357, 1.2988, 0.0485, 0.9270], + device='cuda:0'), covar=tensor([0.1989, 0.1588, 0.0544, 0.0981, 0.2997, 0.0540, 0.2556, 0.1381], + device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0181, 0.0113, 0.0213, 0.0258, 0.0117, 0.0163, 0.0178], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 18:34:54,892 INFO [train.py:901] (0/4) Epoch 16, batch 5200, loss[loss=0.1809, simple_loss=0.2728, pruned_loss=0.04445, over 8364.00 frames. ], tot_loss[loss=0.218, simple_loss=0.2971, pruned_loss=0.06944, over 1610953.88 frames. ], batch size: 24, lr: 4.74e-03, grad_scale: 8.0 +2023-02-06 18:35:03,412 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126458.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:35:10,024 INFO [optim.py:369] (0/4) 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 18:35:10,196 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=126468.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:35:13,937 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.84 vs. limit=5.0 +2023-02-06 18:35:15,120 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126475.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:35:21,832 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 16, batch 5250, loss[loss=0.2352, simple_loss=0.3021, pruned_loss=0.08414, over 7529.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.2975, pruned_loss=0.0695, over 1611155.80 frames. ], batch size: 18, lr: 4.74e-03, grad_scale: 8.0 +2023-02-06 18:35:39,842 WARNING [train.py:1067] (0/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-06 18:36:05,600 INFO [train.py:901] (0/4) Epoch 16, batch 5300, loss[loss=0.2091, simple_loss=0.3017, pruned_loss=0.05821, over 8581.00 frames. ], tot_loss[loss=0.2177, simple_loss=0.2973, pruned_loss=0.06903, over 1614524.12 frames. ], batch size: 31, lr: 4.74e-03, grad_scale: 8.0 +2023-02-06 18:36:20,896 INFO [optim.py:369] (0/4) 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 18:36:37,688 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1866, 2.5240, 2.8179, 1.6711, 3.1131, 1.7302, 1.3643, 2.1862], + device='cuda:0'), covar=tensor([0.0815, 0.0352, 0.0236, 0.0671, 0.0410, 0.0860, 0.0821, 0.0469], + device='cuda:0'), in_proj_covar=tensor([0.0432, 0.0370, 0.0317, 0.0423, 0.0356, 0.0515, 0.0374, 0.0393], + device='cuda:0'), out_proj_covar=tensor([1.1843e-04, 9.8836e-05, 8.4163e-05, 1.1357e-04, 9.5914e-05, 1.4906e-04, + 1.0225e-04, 1.0630e-04], device='cuda:0') +2023-02-06 18:36:41,560 INFO [train.py:901] (0/4) Epoch 16, batch 5350, loss[loss=0.2086, simple_loss=0.2928, pruned_loss=0.06222, over 8348.00 frames. ], tot_loss[loss=0.2198, simple_loss=0.2992, pruned_loss=0.07021, over 1617544.00 frames. ], batch size: 24, lr: 4.73e-03, grad_scale: 8.0 +2023-02-06 18:36:51,955 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.0610, 4.0642, 3.7371, 2.0785, 3.6786, 3.6901, 3.7189, 3.4252], + device='cuda:0'), covar=tensor([0.0901, 0.0662, 0.1060, 0.4417, 0.0878, 0.0977, 0.1238, 0.0902], + device='cuda:0'), in_proj_covar=tensor([0.0494, 0.0409, 0.0412, 0.0510, 0.0403, 0.0407, 0.0399, 0.0352], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 18:36:58,006 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7467, 2.2857, 4.0342, 1.5553, 2.8150, 2.1733, 1.8258, 2.5617], + device='cuda:0'), covar=tensor([0.1797, 0.2257, 0.0910, 0.4027, 0.1718, 0.2869, 0.1973, 0.2503], + device='cuda:0'), in_proj_covar=tensor([0.0500, 0.0553, 0.0540, 0.0611, 0.0627, 0.0566, 0.0498, 0.0617], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 18:37:12,255 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6347, 2.0334, 3.4149, 1.4101, 2.5793, 1.9799, 1.6432, 2.3844], + device='cuda:0'), covar=tensor([0.1849, 0.2285, 0.0895, 0.4278, 0.1641, 0.3049, 0.2128, 0.2345], + device='cuda:0'), in_proj_covar=tensor([0.0501, 0.0554, 0.0541, 0.0612, 0.0628, 0.0567, 0.0498, 0.0618], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 18:37:16,922 INFO [train.py:901] (0/4) Epoch 16, batch 5400, loss[loss=0.2032, simple_loss=0.2794, pruned_loss=0.06355, over 8129.00 frames. ], tot_loss[loss=0.2196, simple_loss=0.299, pruned_loss=0.07011, over 1616278.41 frames. ], batch size: 22, lr: 4.73e-03, grad_scale: 8.0 +2023-02-06 18:37:27,560 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9647, 1.6495, 2.1110, 1.8408, 2.0249, 1.9304, 1.7217, 0.7301], + device='cuda:0'), covar=tensor([0.5105, 0.4428, 0.1674, 0.2916, 0.2036, 0.2688, 0.1822, 0.4476], + device='cuda:0'), in_proj_covar=tensor([0.0908, 0.0916, 0.0756, 0.0892, 0.0958, 0.0841, 0.0717, 0.0795], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 18:37:32,195 INFO [optim.py:369] (0/4) 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 18:37:33,813 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5231, 1.9291, 3.4062, 1.3412, 2.5020, 1.9155, 1.6445, 2.3276], + device='cuda:0'), covar=tensor([0.2007, 0.2449, 0.0771, 0.4603, 0.1831, 0.3140, 0.2264, 0.2461], + device='cuda:0'), in_proj_covar=tensor([0.0504, 0.0557, 0.0544, 0.0615, 0.0631, 0.0570, 0.0501, 0.0622], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 18:37:51,442 INFO [train.py:901] (0/4) Epoch 16, batch 5450, loss[loss=0.246, simple_loss=0.3218, pruned_loss=0.08512, over 8701.00 frames. ], tot_loss[loss=0.2195, simple_loss=0.2985, pruned_loss=0.07021, over 1612038.52 frames. ], batch size: 34, lr: 4.73e-03, grad_scale: 8.0 +2023-02-06 18:37:56,215 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([6.0868, 1.8529, 6.2119, 2.4126, 5.7485, 5.2391, 5.8095, 5.6981], + device='cuda:0'), covar=tensor([0.0366, 0.4137, 0.0333, 0.3102, 0.0748, 0.0809, 0.0387, 0.0412], + device='cuda:0'), in_proj_covar=tensor([0.0555, 0.0613, 0.0633, 0.0580, 0.0659, 0.0565, 0.0556, 0.0621], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 18:38:00,362 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.2684, 3.1733, 2.9910, 1.6723, 2.9075, 2.8733, 2.8787, 2.7544], + device='cuda:0'), covar=tensor([0.1183, 0.0856, 0.1319, 0.4269, 0.1162, 0.1136, 0.1591, 0.1125], + device='cuda:0'), in_proj_covar=tensor([0.0493, 0.0407, 0.0410, 0.0507, 0.0401, 0.0408, 0.0398, 0.0351], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 18:38:17,596 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126731.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:38:17,746 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.57 vs. limit=5.0 +2023-02-06 18:38:22,997 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4998, 2.5670, 1.8079, 2.2163, 2.1792, 1.5904, 2.0009, 2.1332], + device='cuda:0'), covar=tensor([0.1562, 0.0419, 0.1220, 0.0669, 0.0763, 0.1529, 0.1041, 0.0996], + device='cuda:0'), in_proj_covar=tensor([0.0350, 0.0232, 0.0323, 0.0301, 0.0299, 0.0331, 0.0343, 0.0315], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 18:38:24,939 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 16, batch 5500, loss[loss=0.198, simple_loss=0.2661, pruned_loss=0.06493, over 7693.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.2977, pruned_loss=0.07057, over 1602994.52 frames. ], batch size: 18, lr: 4.73e-03, grad_scale: 8.0 +2023-02-06 18:38:28,988 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-06 18:38:35,369 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=126756.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:38:38,179 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7717, 1.6805, 2.4323, 1.6465, 1.2508, 2.2953, 0.5161, 1.4305], + device='cuda:0'), covar=tensor([0.1736, 0.1295, 0.0264, 0.1333, 0.2962, 0.0370, 0.2380, 0.1333], + device='cuda:0'), in_proj_covar=tensor([0.0177, 0.0183, 0.0115, 0.0214, 0.0262, 0.0119, 0.0166, 0.0179], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 18:38:44,227 INFO [optim.py:369] (0/4) 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 18:39:00,749 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 +2023-02-06 18:39:02,248 INFO [train.py:901] (0/4) Epoch 16, batch 5550, loss[loss=0.2239, simple_loss=0.3115, pruned_loss=0.06811, over 8465.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.2969, pruned_loss=0.06967, over 1605909.81 frames. ], batch size: 27, lr: 4.73e-03, grad_scale: 4.0 +2023-02-06 18:39:05,931 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.1466, 1.4802, 4.2834, 1.6919, 3.8039, 3.5767, 3.8989, 3.7666], + device='cuda:0'), covar=tensor([0.0515, 0.4225, 0.0587, 0.3585, 0.1023, 0.0914, 0.0580, 0.0642], + device='cuda:0'), in_proj_covar=tensor([0.0554, 0.0611, 0.0632, 0.0578, 0.0657, 0.0563, 0.0552, 0.0619], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 18:39:13,432 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=126812.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:39:16,822 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3747, 1.4568, 1.4103, 1.8623, 0.7190, 1.1993, 1.3734, 1.4883], + device='cuda:0'), covar=tensor([0.0862, 0.0763, 0.0963, 0.0470, 0.1113, 0.1422, 0.0671, 0.0681], + device='cuda:0'), in_proj_covar=tensor([0.0230, 0.0201, 0.0248, 0.0212, 0.0209, 0.0248, 0.0254, 0.0212], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 18:39:22,292 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2407, 1.9610, 2.6840, 2.1883, 2.6810, 2.1679, 1.8597, 1.3354], + device='cuda:0'), covar=tensor([0.4563, 0.4126, 0.1523, 0.3241, 0.2157, 0.2587, 0.1826, 0.4682], + device='cuda:0'), in_proj_covar=tensor([0.0914, 0.0923, 0.0759, 0.0899, 0.0963, 0.0846, 0.0721, 0.0799], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 18:39:30,339 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 16, batch 5600, loss[loss=0.2393, simple_loss=0.3153, pruned_loss=0.08166, over 7111.00 frames. ], tot_loss[loss=0.2173, simple_loss=0.2963, pruned_loss=0.06918, over 1606066.15 frames. ], batch size: 72, lr: 4.73e-03, grad_scale: 8.0 +2023-02-06 18:39:45,803 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 18:40:12,818 INFO [train.py:901] (0/4) Epoch 16, batch 5650, loss[loss=0.2545, simple_loss=0.3293, pruned_loss=0.08985, over 8182.00 frames. ], tot_loss[loss=0.2169, simple_loss=0.2962, pruned_loss=0.06882, over 1609161.33 frames. ], batch size: 23, lr: 4.73e-03, grad_scale: 8.0 +2023-02-06 18:40:26,274 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.1838, 1.7152, 4.3307, 1.5852, 3.8522, 3.5681, 3.9275, 3.7743], + device='cuda:0'), covar=tensor([0.0522, 0.3778, 0.0494, 0.3700, 0.0968, 0.0916, 0.0521, 0.0618], + device='cuda:0'), in_proj_covar=tensor([0.0554, 0.0610, 0.0631, 0.0580, 0.0658, 0.0562, 0.0553, 0.0618], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 18:40:33,383 WARNING [train.py:1067] (0/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-06 18:40:33,510 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126927.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:40:48,557 INFO [train.py:901] (0/4) Epoch 16, batch 5700, loss[loss=0.2188, simple_loss=0.2913, pruned_loss=0.07317, over 7925.00 frames. ], tot_loss[loss=0.218, simple_loss=0.2971, pruned_loss=0.06945, over 1610080.95 frames. ], batch size: 20, lr: 4.73e-03, grad_scale: 8.0 +2023-02-06 18:41:04,179 INFO [optim.py:369] (0/4) 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 18:41:22,813 INFO [train.py:901] (0/4) Epoch 16, batch 5750, loss[loss=0.189, simple_loss=0.2617, pruned_loss=0.05811, over 7788.00 frames. ], tot_loss[loss=0.2184, simple_loss=0.2977, pruned_loss=0.06959, over 1611511.72 frames. ], batch size: 19, lr: 4.73e-03, grad_scale: 8.0 +2023-02-06 18:41:34,413 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2100, 1.0983, 1.2960, 1.1083, 0.9423, 1.3251, 0.0853, 0.8966], + device='cuda:0'), covar=tensor([0.1659, 0.1444, 0.0463, 0.0779, 0.3055, 0.0505, 0.2374, 0.1387], + device='cuda:0'), in_proj_covar=tensor([0.0177, 0.0184, 0.0116, 0.0216, 0.0262, 0.0120, 0.0168, 0.0181], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 18:41:39,578 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-06 18:41:49,998 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7926, 2.4267, 4.4653, 1.5596, 3.3532, 2.4498, 1.8891, 2.9849], + device='cuda:0'), covar=tensor([0.1709, 0.2245, 0.0678, 0.4094, 0.1486, 0.2675, 0.1997, 0.2260], + device='cuda:0'), in_proj_covar=tensor([0.0503, 0.0553, 0.0541, 0.0611, 0.0625, 0.0564, 0.0499, 0.0617], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 18:41:51,104 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 +2023-02-06 18:41:56,554 INFO [train.py:901] (0/4) Epoch 16, batch 5800, loss[loss=0.1905, simple_loss=0.2593, pruned_loss=0.06087, over 7218.00 frames. ], tot_loss[loss=0.2179, simple_loss=0.2971, pruned_loss=0.06936, over 1611115.42 frames. ], batch size: 16, lr: 4.73e-03, grad_scale: 8.0 +2023-02-06 18:42:14,357 INFO [optim.py:369] (0/4) 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 18:42:33,212 INFO [train.py:901] (0/4) Epoch 16, batch 5850, loss[loss=0.2078, simple_loss=0.2944, pruned_loss=0.06061, over 8251.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2961, pruned_loss=0.06841, over 1612574.14 frames. ], batch size: 22, lr: 4.73e-03, grad_scale: 8.0 +2023-02-06 18:42:45,191 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127113.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:42:47,577 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 +2023-02-06 18:43:02,056 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127138.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:43:07,131 INFO [train.py:901] (0/4) Epoch 16, batch 5900, loss[loss=0.2197, simple_loss=0.2951, pruned_loss=0.07211, over 8137.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.2953, pruned_loss=0.06808, over 1615590.83 frames. ], batch size: 22, lr: 4.72e-03, grad_scale: 8.0 +2023-02-06 18:43:10,073 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.3663, 1.4262, 4.5481, 1.7833, 4.0320, 3.8313, 4.1257, 4.0218], + device='cuda:0'), covar=tensor([0.0516, 0.4630, 0.0480, 0.3708, 0.1117, 0.0878, 0.0561, 0.0641], + device='cuda:0'), in_proj_covar=tensor([0.0552, 0.0609, 0.0631, 0.0579, 0.0659, 0.0561, 0.0553, 0.0618], + device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 18:43:22,997 INFO [optim.py:369] (0/4) 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 18:43:30,606 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=127178.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:43:34,101 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127183.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:43:42,722 INFO [train.py:901] (0/4) Epoch 16, batch 5950, loss[loss=0.2779, simple_loss=0.3387, pruned_loss=0.1086, over 6675.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2947, pruned_loss=0.06777, over 1609474.49 frames. ], batch size: 72, lr: 4.72e-03, grad_scale: 8.0 +2023-02-06 18:43:51,272 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127208.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:43:56,247 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.01 vs. limit=5.0 +2023-02-06 18:44:02,355 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5211, 1.9056, 2.0069, 1.0405, 2.1451, 1.4795, 0.5621, 1.8404], + device='cuda:0'), covar=tensor([0.0497, 0.0328, 0.0231, 0.0546, 0.0318, 0.0742, 0.0713, 0.0259], + device='cuda:0'), in_proj_covar=tensor([0.0429, 0.0368, 0.0316, 0.0422, 0.0355, 0.0518, 0.0373, 0.0394], + device='cuda:0'), out_proj_covar=tensor([1.1757e-04, 9.8481e-05, 8.4071e-05, 1.1340e-04, 9.5717e-05, 1.5004e-04, + 1.0199e-04, 1.0655e-04], device='cuda:0') +2023-02-06 18:44:17,703 INFO [train.py:901] (0/4) Epoch 16, batch 6000, loss[loss=0.1759, simple_loss=0.2548, pruned_loss=0.04851, over 7287.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.2952, pruned_loss=0.06787, over 1608359.78 frames. ], batch size: 16, lr: 4.72e-03, grad_scale: 8.0 +2023-02-06 18:44:17,704 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 18:44:29,966 INFO [train.py:935] (0/4) Epoch 16, validation: loss=0.1793, simple_loss=0.2799, pruned_loss=0.03935, over 944034.00 frames. +2023-02-06 18:44:29,967 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-06 18:44:44,479 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 18:45:01,805 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=127293.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:45:01,838 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5548, 1.9657, 2.0929, 1.0695, 2.1062, 1.4912, 0.6055, 1.9309], + device='cuda:0'), covar=tensor([0.0622, 0.0355, 0.0272, 0.0594, 0.0454, 0.0808, 0.0817, 0.0272], + device='cuda:0'), in_proj_covar=tensor([0.0424, 0.0364, 0.0314, 0.0418, 0.0351, 0.0512, 0.0368, 0.0391], + device='cuda:0'), out_proj_covar=tensor([1.1621e-04, 9.7185e-05, 8.3308e-05, 1.1233e-04, 9.4595e-05, 1.4822e-04, + 1.0077e-04, 1.0562e-04], device='cuda:0') +2023-02-06 18:45:03,659 INFO [train.py:901] (0/4) Epoch 16, batch 6050, loss[loss=0.2404, simple_loss=0.3093, pruned_loss=0.08572, over 7545.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.2952, pruned_loss=0.06758, over 1609519.79 frames. ], batch size: 18, lr: 4.72e-03, grad_scale: 8.0 +2023-02-06 18:45:39,307 INFO [train.py:901] (0/4) Epoch 16, batch 6100, loss[loss=0.2258, simple_loss=0.3123, pruned_loss=0.06962, over 8634.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.2951, pruned_loss=0.06731, over 1615890.11 frames. ], batch size: 49, lr: 4.72e-03, grad_scale: 8.0 +2023-02-06 18:45:54,962 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8873, 1.5084, 3.1999, 1.4313, 2.3178, 3.4991, 3.5518, 3.0154], + device='cuda:0'), covar=tensor([0.1176, 0.1802, 0.0424, 0.2144, 0.1110, 0.0251, 0.0615, 0.0589], + device='cuda:0'), in_proj_covar=tensor([0.0281, 0.0313, 0.0275, 0.0302, 0.0294, 0.0251, 0.0384, 0.0298], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-06 18:45:55,484 INFO [optim.py:369] (0/4) 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 18:46:04,777 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.86 vs. limit=5.0 +2023-02-06 18:46:06,187 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 +2023-02-06 18:46:09,094 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-06 18:46:13,720 INFO [train.py:901] (0/4) Epoch 16, batch 6150, loss[loss=0.2291, simple_loss=0.2973, pruned_loss=0.08044, over 8079.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.2948, pruned_loss=0.06732, over 1612772.23 frames. ], batch size: 21, lr: 4.72e-03, grad_scale: 8.0 +2023-02-06 18:46:15,216 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7149, 1.9016, 1.6322, 2.2585, 1.0817, 1.4143, 1.7152, 1.9507], + device='cuda:0'), covar=tensor([0.0813, 0.0708, 0.0974, 0.0506, 0.1119, 0.1411, 0.0839, 0.0723], + device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0202, 0.0247, 0.0211, 0.0207, 0.0247, 0.0253, 0.0211], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 18:46:15,286 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0595, 1.8648, 2.4322, 1.9907, 2.3217, 2.0904, 1.8156, 1.1140], + device='cuda:0'), covar=tensor([0.4634, 0.4101, 0.1469, 0.2968, 0.2002, 0.2480, 0.1751, 0.4365], + device='cuda:0'), in_proj_covar=tensor([0.0911, 0.0923, 0.0766, 0.0897, 0.0962, 0.0845, 0.0723, 0.0796], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 18:46:48,829 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127445.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 18:46:49,325 INFO [train.py:901] (0/4) Epoch 16, batch 6200, loss[loss=0.227, simple_loss=0.2996, pruned_loss=0.07718, over 7920.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.2954, pruned_loss=0.06804, over 1612696.00 frames. ], batch size: 20, lr: 4.72e-03, grad_scale: 8.0 +2023-02-06 18:47:04,240 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3656, 2.8444, 2.4567, 3.9437, 1.8817, 2.1431, 2.6190, 3.0352], + device='cuda:0'), covar=tensor([0.0765, 0.0775, 0.0815, 0.0220, 0.1046, 0.1231, 0.0892, 0.0733], + device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0201, 0.0247, 0.0211, 0.0207, 0.0246, 0.0253, 0.0211], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 18:47:04,711 INFO [optim.py:369] (0/4) 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 18:47:08,571 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.08 vs. limit=5.0 +2023-02-06 18:47:23,438 INFO [train.py:901] (0/4) Epoch 16, batch 6250, loss[loss=0.2386, simple_loss=0.3164, pruned_loss=0.0804, over 8134.00 frames. ], tot_loss[loss=0.2169, simple_loss=0.2967, pruned_loss=0.06854, over 1616395.11 frames. ], batch size: 22, lr: 4.72e-03, grad_scale: 8.0 +2023-02-06 18:47:57,838 INFO [train.py:901] (0/4) Epoch 16, batch 6300, loss[loss=0.2593, simple_loss=0.3329, pruned_loss=0.09288, over 8188.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.2976, pruned_loss=0.06932, over 1619436.77 frames. ], batch size: 23, lr: 4.72e-03, grad_scale: 8.0 +2023-02-06 18:47:58,586 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127547.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:47:59,972 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127549.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:48:14,536 INFO [optim.py:369] (0/4) 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 18:48:14,709 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0575, 2.2069, 1.8455, 2.8867, 1.2367, 1.5243, 1.8899, 2.2592], + device='cuda:0'), covar=tensor([0.0734, 0.0840, 0.1024, 0.0373, 0.1232, 0.1517, 0.1022, 0.0822], + device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0203, 0.0250, 0.0213, 0.0210, 0.0250, 0.0256, 0.0213], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 18:48:17,987 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127574.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:48:32,718 INFO [train.py:901] (0/4) Epoch 16, batch 6350, loss[loss=0.238, simple_loss=0.322, pruned_loss=0.07696, over 8295.00 frames. ], tot_loss[loss=0.2187, simple_loss=0.2973, pruned_loss=0.07007, over 1614587.85 frames. ], batch size: 23, lr: 4.72e-03, grad_scale: 8.0 +2023-02-06 18:48:43,693 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=127611.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:48:45,101 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.9779, 1.6665, 1.9253, 1.6731, 1.0210, 1.6509, 2.2211, 2.4728], + device='cuda:0'), covar=tensor([0.0428, 0.1245, 0.1648, 0.1419, 0.0598, 0.1526, 0.0618, 0.0548], + device='cuda:0'), in_proj_covar=tensor([0.0096, 0.0152, 0.0190, 0.0156, 0.0101, 0.0162, 0.0114, 0.0140], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 18:49:07,008 INFO [train.py:901] (0/4) Epoch 16, batch 6400, loss[loss=0.2595, simple_loss=0.3296, pruned_loss=0.09472, over 8513.00 frames. ], tot_loss[loss=0.2188, simple_loss=0.297, pruned_loss=0.07027, over 1611384.44 frames. ], batch size: 26, lr: 4.72e-03, grad_scale: 8.0 +2023-02-06 18:49:24,184 INFO [optim.py:369] (0/4) 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 18:49:25,915 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.59 vs. limit=5.0 +2023-02-06 18:49:32,551 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127680.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:49:34,762 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-06 18:49:43,231 INFO [train.py:901] (0/4) Epoch 16, batch 6450, loss[loss=0.2141, simple_loss=0.272, pruned_loss=0.0781, over 7531.00 frames. ], tot_loss[loss=0.2185, simple_loss=0.2971, pruned_loss=0.06992, over 1613423.37 frames. ], batch size: 18, lr: 4.71e-03, grad_scale: 8.0 +2023-02-06 18:50:04,136 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=127726.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:50:15,490 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.47 vs. limit=5.0 +2023-02-06 18:50:17,017 INFO [train.py:901] (0/4) Epoch 16, batch 6500, loss[loss=0.2897, simple_loss=0.3514, pruned_loss=0.1139, over 8752.00 frames. ], tot_loss[loss=0.2185, simple_loss=0.2973, pruned_loss=0.0699, over 1616465.46 frames. ], batch size: 30, lr: 4.71e-03, grad_scale: 8.0 +2023-02-06 18:50:32,623 INFO [optim.py:369] (0/4) 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 18:50:48,407 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=127789.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 18:50:52,951 INFO [train.py:901] (0/4) Epoch 16, batch 6550, loss[loss=0.2127, simple_loss=0.2782, pruned_loss=0.07364, over 7808.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.2969, pruned_loss=0.06966, over 1618741.02 frames. ], batch size: 20, lr: 4.71e-03, grad_scale: 8.0 +2023-02-06 18:50:53,780 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127797.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:50:59,469 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4385, 2.0366, 4.3904, 1.3755, 2.9596, 2.0880, 1.4640, 2.8093], + device='cuda:0'), covar=tensor([0.2256, 0.2744, 0.0712, 0.4892, 0.1821, 0.3294, 0.2522, 0.2619], + device='cuda:0'), in_proj_covar=tensor([0.0505, 0.0559, 0.0544, 0.0612, 0.0626, 0.0567, 0.0503, 0.0618], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 18:51:17,242 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-06 18:51:27,456 INFO [train.py:901] (0/4) Epoch 16, batch 6600, loss[loss=0.1888, simple_loss=0.2633, pruned_loss=0.05715, over 7704.00 frames. ], tot_loss[loss=0.2192, simple_loss=0.298, pruned_loss=0.07018, over 1615130.22 frames. ], batch size: 18, lr: 4.71e-03, grad_scale: 8.0 +2023-02-06 18:51:30,345 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0427, 1.6074, 1.3517, 1.5288, 1.3889, 1.2407, 1.1827, 1.3166], + device='cuda:0'), covar=tensor([0.1029, 0.0437, 0.1129, 0.0533, 0.0704, 0.1322, 0.0920, 0.0738], + device='cuda:0'), in_proj_covar=tensor([0.0345, 0.0232, 0.0322, 0.0299, 0.0298, 0.0330, 0.0339, 0.0311], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 18:51:36,809 WARNING [train.py:1067] (0/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-06 18:51:38,966 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6652, 1.8407, 1.6157, 2.2949, 1.0071, 1.3227, 1.6341, 1.7861], + device='cuda:0'), covar=tensor([0.0878, 0.0855, 0.1053, 0.0440, 0.1145, 0.1533, 0.0939, 0.0956], + device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0203, 0.0249, 0.0213, 0.0211, 0.0250, 0.0256, 0.0214], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 18:51:42,287 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127868.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:51:42,775 INFO [optim.py:369] (0/4) 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 18:51:57,071 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127890.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:51:57,604 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=127891.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:52:00,746 INFO [train.py:901] (0/4) Epoch 16, batch 6650, loss[loss=0.2105, simple_loss=0.2923, pruned_loss=0.06436, over 7984.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.297, pruned_loss=0.06963, over 1612661.66 frames. ], batch size: 21, lr: 4.71e-03, grad_scale: 8.0 +2023-02-06 18:52:07,580 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=127904.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 18:52:28,264 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.0642, 1.1811, 3.1995, 1.0781, 2.7826, 2.6718, 2.9163, 2.8132], + device='cuda:0'), covar=tensor([0.0801, 0.4286, 0.0902, 0.4029, 0.1479, 0.1095, 0.0798, 0.0956], + device='cuda:0'), in_proj_covar=tensor([0.0566, 0.0622, 0.0645, 0.0592, 0.0672, 0.0572, 0.0567, 0.0632], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 18:52:36,163 INFO [train.py:901] (0/4) Epoch 16, batch 6700, loss[loss=0.2088, simple_loss=0.2853, pruned_loss=0.06616, over 7967.00 frames. ], tot_loss[loss=0.2169, simple_loss=0.2962, pruned_loss=0.06883, over 1614345.43 frames. ], batch size: 21, lr: 4.71e-03, grad_scale: 8.0 +2023-02-06 18:52:52,425 INFO [optim.py:369] (0/4) 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 18:53:01,215 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 16, batch 6750, loss[loss=0.2397, simple_loss=0.3162, pruned_loss=0.0816, over 8619.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.2965, pruned_loss=0.06915, over 1612796.33 frames. ], batch size: 39, lr: 4.71e-03, grad_scale: 8.0 +2023-02-06 18:53:13,303 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-128000.pt +2023-02-06 18:53:18,491 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128006.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:53:19,157 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128007.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:53:32,876 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=128024.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:53:47,638 INFO [train.py:901] (0/4) Epoch 16, batch 6800, loss[loss=0.2015, simple_loss=0.281, pruned_loss=0.06096, over 7810.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.2967, pruned_loss=0.06941, over 1611272.37 frames. ], batch size: 20, lr: 4.71e-03, grad_scale: 8.0 +2023-02-06 18:53:51,040 WARNING [train.py:1067] (0/4) 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] (0/4) 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 18:54:22,239 INFO [train.py:901] (0/4) Epoch 16, batch 6850, loss[loss=0.1984, simple_loss=0.2933, pruned_loss=0.05175, over 8036.00 frames. ], tot_loss[loss=0.2187, simple_loss=0.2982, pruned_loss=0.06956, over 1615869.12 frames. ], batch size: 22, lr: 4.71e-03, grad_scale: 8.0 +2023-02-06 18:54:36,134 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9761, 2.1411, 1.7762, 2.7223, 1.2196, 1.4965, 1.9715, 2.1693], + device='cuda:0'), covar=tensor([0.0769, 0.0836, 0.0969, 0.0403, 0.1134, 0.1481, 0.0986, 0.0739], + device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0202, 0.0249, 0.0212, 0.0209, 0.0249, 0.0253, 0.0212], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 18:54:40,681 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-06 18:54:50,127 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 +2023-02-06 18:54:53,400 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128139.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:54:54,725 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=128141.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:54:58,091 INFO [train.py:901] (0/4) Epoch 16, batch 6900, loss[loss=0.2203, simple_loss=0.2951, pruned_loss=0.07274, over 8035.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.2977, pruned_loss=0.06926, over 1615154.57 frames. ], batch size: 22, lr: 4.71e-03, grad_scale: 8.0 +2023-02-06 18:54:58,582 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 +2023-02-06 18:55:06,442 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 +2023-02-06 18:55:08,151 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=128160.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 18:55:14,251 INFO [optim.py:369] (0/4) 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 18:55:25,853 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128185.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 18:55:32,787 INFO [train.py:901] (0/4) Epoch 16, batch 6950, loss[loss=0.2299, simple_loss=0.3093, pruned_loss=0.07523, over 8492.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.297, pruned_loss=0.06893, over 1612049.07 frames. ], batch size: 29, lr: 4.71e-03, grad_scale: 8.0 +2023-02-06 18:55:43,515 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=128212.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:55:48,025 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-06 18:55:58,347 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=128234.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:56:05,943 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.8688, 3.7499, 3.4788, 1.8489, 3.3763, 3.4804, 3.3530, 3.3029], + device='cuda:0'), covar=tensor([0.0800, 0.0647, 0.1095, 0.4045, 0.0874, 0.0890, 0.1444, 0.0780], + device='cuda:0'), in_proj_covar=tensor([0.0492, 0.0406, 0.0408, 0.0505, 0.0396, 0.0408, 0.0400, 0.0352], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 18:56:07,210 INFO [train.py:901] (0/4) Epoch 16, batch 7000, loss[loss=0.1918, simple_loss=0.2754, pruned_loss=0.05412, over 7973.00 frames. ], tot_loss[loss=0.2175, simple_loss=0.2967, pruned_loss=0.06917, over 1605294.55 frames. ], batch size: 21, lr: 4.70e-03, grad_scale: 8.0 +2023-02-06 18:56:15,592 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128256.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:56:19,637 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 18:56:36,181 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 16, batch 7050, loss[loss=0.2092, simple_loss=0.2827, pruned_loss=0.06782, over 8292.00 frames. ], tot_loss[loss=0.217, simple_loss=0.2961, pruned_loss=0.06899, over 1605319.78 frames. ], batch size: 23, lr: 4.70e-03, grad_scale: 8.0 +2023-02-06 18:57:03,873 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128338.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:57:16,462 INFO [train.py:901] (0/4) Epoch 16, batch 7100, loss[loss=0.1891, simple_loss=0.2752, pruned_loss=0.05155, over 8523.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.2968, pruned_loss=0.06939, over 1605503.23 frames. ], batch size: 39, lr: 4.70e-03, grad_scale: 8.0 +2023-02-06 18:57:18,638 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128349.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:57:33,895 INFO [optim.py:369] (0/4) 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 18:57:52,015 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 16, batch 7150, loss[loss=0.2569, simple_loss=0.318, pruned_loss=0.09784, over 6699.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.2972, pruned_loss=0.06917, over 1606284.21 frames. ], batch size: 72, lr: 4.70e-03, grad_scale: 8.0 +2023-02-06 18:58:00,908 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7489, 1.7888, 2.5683, 1.6264, 1.2643, 2.4917, 0.4676, 1.4395], + device='cuda:0'), covar=tensor([0.2152, 0.1457, 0.0292, 0.1551, 0.3209, 0.0340, 0.2538, 0.1669], + device='cuda:0'), in_proj_covar=tensor([0.0176, 0.0185, 0.0114, 0.0215, 0.0260, 0.0120, 0.0165, 0.0180], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 18:58:09,831 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128420.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:58:13,952 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6711, 1.9383, 2.0756, 1.2816, 2.1703, 1.5682, 0.6093, 1.8326], + device='cuda:0'), covar=tensor([0.0444, 0.0297, 0.0226, 0.0448, 0.0341, 0.0757, 0.0731, 0.0238], + device='cuda:0'), in_proj_covar=tensor([0.0427, 0.0368, 0.0316, 0.0425, 0.0353, 0.0515, 0.0374, 0.0397], + device='cuda:0'), out_proj_covar=tensor([1.1705e-04, 9.8409e-05, 8.3824e-05, 1.1425e-04, 9.5101e-05, 1.4893e-04, + 1.0248e-04, 1.0749e-04], device='cuda:0') +2023-02-06 18:58:17,204 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 16, batch 7200, loss[loss=0.2506, simple_loss=0.3251, pruned_loss=0.08798, over 8460.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.2971, pruned_loss=0.0693, over 1607243.00 frames. ], batch size: 29, lr: 4.70e-03, grad_scale: 8.0 +2023-02-06 18:58:42,533 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 16, batch 7250, loss[loss=0.2126, simple_loss=0.2949, pruned_loss=0.06515, over 8103.00 frames. ], tot_loss[loss=0.2187, simple_loss=0.2979, pruned_loss=0.06974, over 1605847.20 frames. ], batch size: 23, lr: 4.70e-03, grad_scale: 8.0 +2023-02-06 18:59:13,772 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=128512.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:59:31,355 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128537.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 18:59:37,226 INFO [train.py:901] (0/4) Epoch 16, batch 7300, loss[loss=0.1756, simple_loss=0.2617, pruned_loss=0.04478, over 8141.00 frames. ], tot_loss[loss=0.2175, simple_loss=0.2966, pruned_loss=0.06926, over 1608854.20 frames. ], batch size: 22, lr: 4.70e-03, grad_scale: 8.0 +2023-02-06 18:59:37,480 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4638, 2.1279, 3.1976, 2.5280, 3.1085, 2.3599, 2.0746, 1.8216], + device='cuda:0'), covar=tensor([0.4912, 0.5024, 0.1725, 0.3408, 0.2403, 0.2781, 0.1863, 0.5343], + device='cuda:0'), in_proj_covar=tensor([0.0911, 0.0924, 0.0764, 0.0896, 0.0961, 0.0848, 0.0723, 0.0797], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 18:59:45,716 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 +2023-02-06 18:59:51,672 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.74 vs. limit=5.0 +2023-02-06 18:59:52,609 INFO [optim.py:369] (0/4) 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 19:00:02,308 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=128583.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:00:12,491 INFO [train.py:901] (0/4) Epoch 16, batch 7350, loss[loss=0.2889, simple_loss=0.3437, pruned_loss=0.117, over 6850.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.2971, pruned_loss=0.06976, over 1610716.78 frames. ], batch size: 72, lr: 4.70e-03, grad_scale: 8.0 +2023-02-06 19:00:18,850 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2007, 1.9686, 2.7302, 2.2320, 2.6803, 2.1732, 1.8685, 1.4456], + device='cuda:0'), covar=tensor([0.5074, 0.4883, 0.1778, 0.3337, 0.2349, 0.2668, 0.1878, 0.5089], + device='cuda:0'), in_proj_covar=tensor([0.0916, 0.0927, 0.0768, 0.0903, 0.0965, 0.0853, 0.0727, 0.0801], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 19:00:19,503 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=128605.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:00:21,501 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128608.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:00:31,426 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-06 19:00:36,120 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 16, batch 7400, loss[loss=0.2201, simple_loss=0.2835, pruned_loss=0.07835, over 7689.00 frames. ], tot_loss[loss=0.217, simple_loss=0.296, pruned_loss=0.06897, over 1607181.74 frames. ], batch size: 18, lr: 4.70e-03, grad_scale: 8.0 +2023-02-06 19:00:49,569 WARNING [train.py:1067] (0/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-06 19:01:02,831 INFO [optim.py:369] (0/4) 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 19:01:04,782 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 +2023-02-06 19:01:11,761 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 16, batch 7450, loss[loss=0.1979, simple_loss=0.285, pruned_loss=0.05543, over 8295.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.2967, pruned_loss=0.06944, over 1607048.41 frames. ], batch size: 23, lr: 4.70e-03, grad_scale: 8.0 +2023-02-06 19:01:30,642 WARNING [train.py:1067] (0/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-06 19:01:56,804 INFO [train.py:901] (0/4) Epoch 16, batch 7500, loss[loss=0.1837, simple_loss=0.2739, pruned_loss=0.04674, over 7651.00 frames. ], tot_loss[loss=0.2172, simple_loss=0.2962, pruned_loss=0.06907, over 1611095.93 frames. ], batch size: 19, lr: 4.70e-03, grad_scale: 8.0 +2023-02-06 19:02:13,129 INFO [optim.py:369] (0/4) 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 19:02:17,171 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=128775.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:02:31,136 INFO [train.py:901] (0/4) Epoch 16, batch 7550, loss[loss=0.2025, simple_loss=0.2888, pruned_loss=0.05812, over 8024.00 frames. ], tot_loss[loss=0.2173, simple_loss=0.2962, pruned_loss=0.06916, over 1610634.50 frames. ], batch size: 22, lr: 4.69e-03, grad_scale: 16.0 +2023-02-06 19:02:32,031 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128797.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:02:33,338 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128799.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:03:07,396 INFO [train.py:901] (0/4) Epoch 16, batch 7600, loss[loss=0.2334, simple_loss=0.3174, pruned_loss=0.07467, over 8528.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.2964, pruned_loss=0.06915, over 1613264.37 frames. ], batch size: 28, lr: 4.69e-03, grad_scale: 16.0 +2023-02-06 19:03:16,512 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3216, 2.4484, 1.8913, 2.0941, 1.9388, 1.5706, 1.8756, 1.9721], + device='cuda:0'), covar=tensor([0.1419, 0.0373, 0.1100, 0.0616, 0.0695, 0.1470, 0.0944, 0.0980], + device='cuda:0'), in_proj_covar=tensor([0.0349, 0.0233, 0.0326, 0.0300, 0.0298, 0.0331, 0.0342, 0.0314], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 19:03:21,390 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3277, 1.4893, 1.4252, 1.8608, 0.7792, 1.2113, 1.3293, 1.4785], + device='cuda:0'), covar=tensor([0.0978, 0.0850, 0.1057, 0.0509, 0.1173, 0.1636, 0.0774, 0.0792], + device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0200, 0.0246, 0.0209, 0.0207, 0.0247, 0.0250, 0.0211], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 19:03:22,829 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6114, 1.9535, 1.9688, 1.2158, 2.0614, 1.6071, 0.4484, 1.7899], + device='cuda:0'), covar=tensor([0.0394, 0.0272, 0.0221, 0.0438, 0.0354, 0.0718, 0.0690, 0.0225], + device='cuda:0'), in_proj_covar=tensor([0.0429, 0.0368, 0.0317, 0.0424, 0.0353, 0.0514, 0.0375, 0.0395], + device='cuda:0'), out_proj_covar=tensor([1.1750e-04, 9.8324e-05, 8.4154e-05, 1.1383e-04, 9.4979e-05, 1.4849e-04, + 1.0268e-04, 1.0664e-04], device='cuda:0') +2023-02-06 19:03:23,268 INFO [optim.py:369] (0/4) 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 19:03:38,600 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128890.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:03:42,475 INFO [train.py:901] (0/4) Epoch 16, batch 7650, loss[loss=0.2215, simple_loss=0.312, pruned_loss=0.06549, over 8473.00 frames. ], tot_loss[loss=0.217, simple_loss=0.2962, pruned_loss=0.06885, over 1613424.37 frames. ], batch size: 25, lr: 4.69e-03, grad_scale: 16.0 +2023-02-06 19:03:50,519 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128908.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 19:03:52,627 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3820, 2.0444, 2.8350, 2.3289, 2.8318, 2.3622, 2.0014, 1.4325], + device='cuda:0'), covar=tensor([0.4921, 0.4685, 0.1597, 0.2988, 0.2014, 0.2602, 0.1808, 0.5066], + device='cuda:0'), in_proj_covar=tensor([0.0906, 0.0917, 0.0759, 0.0895, 0.0955, 0.0844, 0.0718, 0.0791], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 19:04:17,494 INFO [train.py:901] (0/4) Epoch 16, batch 7700, loss[loss=0.2132, simple_loss=0.3034, pruned_loss=0.06144, over 8246.00 frames. ], tot_loss[loss=0.217, simple_loss=0.2962, pruned_loss=0.06891, over 1615710.15 frames. ], batch size: 24, lr: 4.69e-03, grad_scale: 8.0 +2023-02-06 19:04:34,543 INFO [optim.py:369] (0/4) 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 19:04:42,146 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-06 19:04:52,876 INFO [train.py:901] (0/4) Epoch 16, batch 7750, loss[loss=0.2455, simple_loss=0.317, pruned_loss=0.08703, over 8696.00 frames. ], tot_loss[loss=0.2165, simple_loss=0.2957, pruned_loss=0.06871, over 1614991.81 frames. ], batch size: 34, lr: 4.69e-03, grad_scale: 8.0 +2023-02-06 19:05:26,240 INFO [train.py:901] (0/4) Epoch 16, batch 7800, loss[loss=0.1985, simple_loss=0.2612, pruned_loss=0.06788, over 7699.00 frames. ], tot_loss[loss=0.2159, simple_loss=0.2956, pruned_loss=0.06813, over 1617099.35 frames. ], batch size: 18, lr: 4.69e-03, grad_scale: 8.0 +2023-02-06 19:05:31,086 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=129053.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:05:41,938 INFO [optim.py:369] (0/4) 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 19:05:48,210 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 16, batch 7850, loss[loss=0.2002, simple_loss=0.2957, pruned_loss=0.05235, over 8335.00 frames. ], tot_loss[loss=0.2173, simple_loss=0.2965, pruned_loss=0.06902, over 1618413.55 frames. ], batch size: 25, lr: 4.69e-03, grad_scale: 8.0 +2023-02-06 19:06:21,688 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5132, 1.4974, 1.8526, 1.4163, 1.1874, 1.8132, 0.1886, 1.2555], + device='cuda:0'), covar=tensor([0.2189, 0.1444, 0.0443, 0.1064, 0.3204, 0.0503, 0.2637, 0.1349], + device='cuda:0'), in_proj_covar=tensor([0.0178, 0.0185, 0.0115, 0.0216, 0.0264, 0.0121, 0.0166, 0.0180], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 19:06:32,322 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 16, batch 7900, loss[loss=0.2161, simple_loss=0.2978, pruned_loss=0.06719, over 8462.00 frames. ], tot_loss[loss=0.2166, simple_loss=0.2959, pruned_loss=0.06865, over 1614106.90 frames. ], batch size: 25, lr: 4.69e-03, grad_scale: 8.0 +2023-02-06 19:06:34,513 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 19:06:51,889 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=129171.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:07:08,449 INFO [train.py:901] (0/4) Epoch 16, batch 7950, loss[loss=0.2943, simple_loss=0.347, pruned_loss=0.1208, over 6745.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.2973, pruned_loss=0.06967, over 1616200.67 frames. ], batch size: 71, lr: 4.69e-03, grad_scale: 8.0 +2023-02-06 19:07:27,495 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2793, 2.0344, 2.9177, 2.3050, 2.6939, 2.3049, 1.9582, 1.5284], + device='cuda:0'), covar=tensor([0.4864, 0.4763, 0.1561, 0.3410, 0.2375, 0.2578, 0.1759, 0.4886], + device='cuda:0'), in_proj_covar=tensor([0.0905, 0.0918, 0.0758, 0.0892, 0.0954, 0.0844, 0.0717, 0.0792], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 19:07:42,917 INFO [train.py:901] (0/4) Epoch 16, batch 8000, loss[loss=0.2111, simple_loss=0.2975, pruned_loss=0.06233, over 8028.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.297, pruned_loss=0.06924, over 1617205.43 frames. ], batch size: 22, lr: 4.69e-03, grad_scale: 8.0 +2023-02-06 19:07:47,177 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=129252.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 19:07:51,178 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 19:08:16,584 INFO [train.py:901] (0/4) Epoch 16, batch 8050, loss[loss=0.1775, simple_loss=0.2651, pruned_loss=0.04498, over 7937.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.2947, pruned_loss=0.06881, over 1595630.15 frames. ], batch size: 20, lr: 4.69e-03, grad_scale: 8.0 +2023-02-06 19:08:40,007 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/epoch-16.pt +2023-02-06 19:08:52,804 WARNING [train.py:1067] (0/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-06 19:08:56,563 INFO [train.py:901] (0/4) Epoch 17, batch 0, loss[loss=0.2397, simple_loss=0.315, pruned_loss=0.08214, over 8472.00 frames. ], tot_loss[loss=0.2397, simple_loss=0.315, pruned_loss=0.08214, over 8472.00 frames. ], batch size: 25, lr: 4.54e-03, grad_scale: 8.0 +2023-02-06 19:08:56,564 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 19:09:04,866 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5004, 1.7277, 2.5627, 1.3157, 1.8867, 1.7808, 1.5535, 1.8138], + device='cuda:0'), covar=tensor([0.1725, 0.2444, 0.0793, 0.4270, 0.1698, 0.3018, 0.2123, 0.2087], + device='cuda:0'), in_proj_covar=tensor([0.0504, 0.0561, 0.0543, 0.0612, 0.0631, 0.0570, 0.0501, 0.0620], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 19:09:07,558 INFO [train.py:935] (0/4) Epoch 17, validation: loss=0.1792, simple_loss=0.2794, pruned_loss=0.03944, over 944034.00 frames. +2023-02-06 19:09:07,559 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-06 19:09:21,171 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-06 19:09:26,471 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.83 vs. limit=5.0 +2023-02-06 19:09:33,851 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=129367.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 19:09:35,634 INFO [optim.py:369] (0/4) 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 19:09:41,816 INFO [train.py:901] (0/4) Epoch 17, batch 50, loss[loss=0.1754, simple_loss=0.251, pruned_loss=0.0499, over 7555.00 frames. ], tot_loss[loss=0.221, simple_loss=0.3001, pruned_loss=0.07089, over 366094.84 frames. ], batch size: 18, lr: 4.54e-03, grad_scale: 8.0 +2023-02-06 19:09:54,005 WARNING [train.py:1067] (0/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-06 19:10:18,443 INFO [train.py:901] (0/4) Epoch 17, batch 100, loss[loss=0.2604, simple_loss=0.3235, pruned_loss=0.09863, over 7048.00 frames. ], tot_loss[loss=0.2196, simple_loss=0.2993, pruned_loss=0.06994, over 646324.58 frames. ], batch size: 71, lr: 4.54e-03, grad_scale: 8.0 +2023-02-06 19:10:18,452 WARNING [train.py:1067] (0/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-06 19:10:19,954 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=129431.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 19:10:28,659 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([6.0469, 1.6653, 6.1843, 2.1993, 5.7142, 5.2718, 5.7477, 5.7150], + device='cuda:0'), covar=tensor([0.0369, 0.4432, 0.0292, 0.3609, 0.0787, 0.0770, 0.0358, 0.0387], + device='cuda:0'), in_proj_covar=tensor([0.0559, 0.0618, 0.0638, 0.0587, 0.0664, 0.0565, 0.0563, 0.0622], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 19:10:31,368 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.3369, 1.4773, 4.5469, 1.7707, 4.0034, 3.8429, 4.1300, 4.0491], + device='cuda:0'), covar=tensor([0.0517, 0.4527, 0.0477, 0.3658, 0.1091, 0.0916, 0.0516, 0.0603], + device='cuda:0'), in_proj_covar=tensor([0.0559, 0.0618, 0.0638, 0.0587, 0.0664, 0.0566, 0.0563, 0.0622], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 19:10:44,180 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6073, 1.8144, 2.6591, 1.4658, 2.2137, 1.8292, 1.6613, 2.0642], + device='cuda:0'), covar=tensor([0.1438, 0.1964, 0.0702, 0.3396, 0.1370, 0.2367, 0.1707, 0.1999], + device='cuda:0'), in_proj_covar=tensor([0.0508, 0.0564, 0.0547, 0.0615, 0.0635, 0.0572, 0.0506, 0.0623], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 19:10:46,022 INFO [optim.py:369] (0/4) 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 19:10:52,175 INFO [train.py:901] (0/4) Epoch 17, batch 150, loss[loss=0.2192, simple_loss=0.3016, pruned_loss=0.06845, over 8657.00 frames. ], tot_loss[loss=0.2218, simple_loss=0.3012, pruned_loss=0.07126, over 859628.51 frames. ], batch size: 34, lr: 4.54e-03, grad_scale: 8.0 +2023-02-06 19:11:18,269 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=129514.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:11:29,026 INFO [train.py:901] (0/4) Epoch 17, batch 200, loss[loss=0.1751, simple_loss=0.2522, pruned_loss=0.04897, over 7420.00 frames. ], tot_loss[loss=0.22, simple_loss=0.2993, pruned_loss=0.07032, over 1027570.09 frames. ], batch size: 17, lr: 4.54e-03, grad_scale: 8.0 +2023-02-06 19:11:36,223 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=129539.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:11:57,078 INFO [optim.py:369] (0/4) 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 19:12:03,435 INFO [train.py:901] (0/4) Epoch 17, batch 250, loss[loss=0.1808, simple_loss=0.2604, pruned_loss=0.05061, over 7977.00 frames. ], tot_loss[loss=0.2193, simple_loss=0.2987, pruned_loss=0.06993, over 1159094.41 frames. ], batch size: 21, lr: 4.54e-03, grad_scale: 8.0 +2023-02-06 19:12:09,633 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-06 19:12:11,839 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7722, 3.6623, 2.2559, 2.5613, 2.5237, 1.9358, 2.4347, 2.8663], + device='cuda:0'), covar=tensor([0.1653, 0.0284, 0.1150, 0.0812, 0.0815, 0.1469, 0.1149, 0.1037], + device='cuda:0'), in_proj_covar=tensor([0.0355, 0.0236, 0.0330, 0.0303, 0.0301, 0.0337, 0.0344, 0.0316], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 19:12:18,377 WARNING [train.py:1067] (0/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-06 19:12:33,720 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=129623.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 19:12:38,226 INFO [train.py:901] (0/4) Epoch 17, batch 300, loss[loss=0.2374, simple_loss=0.3138, pruned_loss=0.08056, over 8035.00 frames. ], tot_loss[loss=0.218, simple_loss=0.2976, pruned_loss=0.06921, over 1267026.53 frames. ], batch size: 22, lr: 4.54e-03, grad_scale: 8.0 +2023-02-06 19:12:39,081 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=129630.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:12:53,503 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=129648.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 19:13:08,146 INFO [optim.py:369] (0/4) 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 19:13:14,341 INFO [train.py:901] (0/4) Epoch 17, batch 350, loss[loss=0.2299, simple_loss=0.3109, pruned_loss=0.0745, over 8500.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.2979, pruned_loss=0.06932, over 1345410.64 frames. ], batch size: 26, lr: 4.54e-03, grad_scale: 8.0 +2023-02-06 19:13:47,838 INFO [train.py:901] (0/4) Epoch 17, batch 400, loss[loss=0.2156, simple_loss=0.293, pruned_loss=0.06912, over 8192.00 frames. ], tot_loss[loss=0.2175, simple_loss=0.2973, pruned_loss=0.06879, over 1404170.04 frames. ], batch size: 23, lr: 4.54e-03, grad_scale: 8.0 +2023-02-06 19:13:57,416 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9401, 2.0798, 1.7756, 2.4932, 1.2151, 1.5846, 1.8312, 2.1733], + device='cuda:0'), covar=tensor([0.0757, 0.0815, 0.0985, 0.0438, 0.1189, 0.1437, 0.0867, 0.0740], + device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0202, 0.0248, 0.0212, 0.0211, 0.0249, 0.0253, 0.0212], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 19:14:18,001 INFO [optim.py:369] (0/4) 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 19:14:21,455 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=129775.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 19:14:24,069 INFO [train.py:901] (0/4) Epoch 17, batch 450, loss[loss=0.2144, simple_loss=0.2949, pruned_loss=0.06699, over 8363.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.2976, pruned_loss=0.06864, over 1453639.38 frames. ], batch size: 24, lr: 4.54e-03, grad_scale: 8.0 +2023-02-06 19:14:44,499 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=129809.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:14:54,733 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.7941, 5.9225, 5.1165, 2.5055, 5.1545, 5.5442, 5.4073, 5.2336], + device='cuda:0'), covar=tensor([0.0514, 0.0378, 0.0960, 0.4216, 0.0696, 0.0852, 0.1039, 0.0624], + device='cuda:0'), in_proj_covar=tensor([0.0505, 0.0414, 0.0416, 0.0512, 0.0406, 0.0416, 0.0404, 0.0358], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 19:14:58,027 INFO [train.py:901] (0/4) Epoch 17, batch 500, loss[loss=0.198, simple_loss=0.2911, pruned_loss=0.05247, over 8107.00 frames. ], tot_loss[loss=0.2173, simple_loss=0.2972, pruned_loss=0.06867, over 1489161.75 frames. ], batch size: 23, lr: 4.53e-03, grad_scale: 8.0 +2023-02-06 19:15:20,979 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.8112, 1.4161, 3.9578, 1.3588, 3.5043, 3.2954, 3.5681, 3.4659], + device='cuda:0'), covar=tensor([0.0617, 0.4401, 0.0618, 0.4144, 0.1249, 0.1130, 0.0695, 0.0774], + device='cuda:0'), in_proj_covar=tensor([0.0561, 0.0620, 0.0643, 0.0590, 0.0669, 0.0574, 0.0569, 0.0628], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 19:15:28,004 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 17, batch 550, loss[loss=0.3088, simple_loss=0.3631, pruned_loss=0.1272, over 7232.00 frames. ], tot_loss[loss=0.2173, simple_loss=0.2975, pruned_loss=0.06853, over 1516461.71 frames. ], batch size: 72, lr: 4.53e-03, grad_scale: 8.0 +2023-02-06 19:15:43,362 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=129890.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 19:16:10,114 INFO [train.py:901] (0/4) Epoch 17, batch 600, loss[loss=0.2519, simple_loss=0.3263, pruned_loss=0.08877, over 8669.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.296, pruned_loss=0.0677, over 1540374.83 frames. ], batch size: 39, lr: 4.53e-03, grad_scale: 8.0 +2023-02-06 19:16:19,705 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-06 19:16:38,503 INFO [optim.py:369] (0/4) 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 19:16:41,330 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=129974.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:16:44,755 INFO [train.py:901] (0/4) Epoch 17, batch 650, loss[loss=0.2017, simple_loss=0.2904, pruned_loss=0.0565, over 8485.00 frames. ], tot_loss[loss=0.2141, simple_loss=0.2947, pruned_loss=0.06678, over 1556403.19 frames. ], batch size: 29, lr: 4.53e-03, grad_scale: 8.0 +2023-02-06 19:17:02,611 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-130000.pt +2023-02-06 19:17:16,433 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130018.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:17:17,837 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130020.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:17:23,793 INFO [train.py:901] (0/4) Epoch 17, batch 700, loss[loss=0.2515, simple_loss=0.3307, pruned_loss=0.08614, over 8610.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.2936, pruned_loss=0.06641, over 1569389.83 frames. ], batch size: 34, lr: 4.53e-03, grad_scale: 8.0 +2023-02-06 19:17:32,190 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1627, 2.3628, 1.8183, 2.8807, 1.3236, 1.6479, 1.8933, 2.3490], + device='cuda:0'), covar=tensor([0.0654, 0.0724, 0.0957, 0.0328, 0.1233, 0.1416, 0.1028, 0.0808], + device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0201, 0.0249, 0.0211, 0.0210, 0.0250, 0.0252, 0.0214], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 19:17:51,866 INFO [optim.py:369] (0/4) 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 19:17:53,079 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.83 vs. limit=5.0 +2023-02-06 19:17:58,270 INFO [train.py:901] (0/4) Epoch 17, batch 750, loss[loss=0.1967, simple_loss=0.2828, pruned_loss=0.05534, over 8302.00 frames. ], tot_loss[loss=0.2137, simple_loss=0.2935, pruned_loss=0.06699, over 1581297.72 frames. ], batch size: 23, lr: 4.53e-03, grad_scale: 8.0 +2023-02-06 19:18:05,545 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130089.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:18:08,269 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-06 19:18:19,413 WARNING [train.py:1067] (0/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-06 19:18:36,034 INFO [train.py:901] (0/4) Epoch 17, batch 800, loss[loss=0.2294, simple_loss=0.3145, pruned_loss=0.07214, over 8331.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.2942, pruned_loss=0.06739, over 1586748.65 frames. ], batch size: 25, lr: 4.53e-03, grad_scale: 8.0 +2023-02-06 19:18:48,055 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130146.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 19:18:52,755 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=130153.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:19:04,225 INFO [optim.py:369] (0/4) 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 19:19:05,153 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130171.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 19:19:10,475 INFO [train.py:901] (0/4) Epoch 17, batch 850, loss[loss=0.1838, simple_loss=0.2576, pruned_loss=0.05499, over 7431.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.295, pruned_loss=0.0677, over 1595321.57 frames. ], batch size: 17, lr: 4.53e-03, grad_scale: 8.0 +2023-02-06 19:19:47,307 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 +2023-02-06 19:19:47,570 INFO [train.py:901] (0/4) Epoch 17, batch 900, loss[loss=0.2205, simple_loss=0.2972, pruned_loss=0.0719, over 8124.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.2958, pruned_loss=0.06773, over 1606436.53 frames. ], batch size: 22, lr: 4.53e-03, grad_scale: 8.0 +2023-02-06 19:20:15,364 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130268.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:20:16,502 INFO [optim.py:369] (0/4) 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 19:20:19,508 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7493, 1.8259, 2.3849, 1.5458, 1.4286, 2.3862, 0.3189, 1.4309], + device='cuda:0'), covar=tensor([0.1983, 0.1169, 0.0308, 0.1362, 0.2694, 0.0366, 0.2454, 0.1470], + device='cuda:0'), in_proj_covar=tensor([0.0177, 0.0184, 0.0115, 0.0217, 0.0262, 0.0120, 0.0166, 0.0178], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 19:20:22,803 INFO [train.py:901] (0/4) Epoch 17, batch 950, loss[loss=0.2304, simple_loss=0.3078, pruned_loss=0.07655, over 8303.00 frames. ], tot_loss[loss=0.2159, simple_loss=0.2958, pruned_loss=0.06796, over 1605627.63 frames. ], batch size: 23, lr: 4.53e-03, grad_scale: 8.0 +2023-02-06 19:20:43,393 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-06 19:20:57,198 INFO [train.py:901] (0/4) Epoch 17, batch 1000, loss[loss=0.2021, simple_loss=0.2897, pruned_loss=0.05727, over 8190.00 frames. ], tot_loss[loss=0.2154, simple_loss=0.2956, pruned_loss=0.0676, over 1609301.82 frames. ], batch size: 23, lr: 4.53e-03, grad_scale: 8.0 +2023-02-06 19:21:04,864 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8223, 2.0242, 1.7480, 2.6053, 1.2712, 1.5375, 1.8448, 2.0694], + device='cuda:0'), covar=tensor([0.0779, 0.0799, 0.1072, 0.0402, 0.1082, 0.1420, 0.0856, 0.0784], + device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0202, 0.0252, 0.0213, 0.0210, 0.0250, 0.0255, 0.0214], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 19:21:09,193 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130345.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:21:20,018 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-06 19:21:21,936 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=130362.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:21:23,300 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=130364.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:21:27,497 INFO [optim.py:369] (0/4) 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 19:21:27,749 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130370.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:21:32,308 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 +2023-02-06 19:21:33,160 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-06 19:21:33,848 INFO [train.py:901] (0/4) Epoch 17, batch 1050, loss[loss=0.194, simple_loss=0.2846, pruned_loss=0.05175, over 8136.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.2965, pruned_loss=0.06795, over 1613061.44 frames. ], batch size: 22, lr: 4.52e-03, grad_scale: 8.0 +2023-02-06 19:21:49,924 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130402.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:22:08,442 INFO [train.py:901] (0/4) Epoch 17, batch 1100, loss[loss=0.1895, simple_loss=0.2774, pruned_loss=0.05083, over 8260.00 frames. ], tot_loss[loss=0.2176, simple_loss=0.2977, pruned_loss=0.06873, over 1619330.83 frames. ], batch size: 24, lr: 4.52e-03, grad_scale: 8.0 +2023-02-06 19:22:23,041 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130450.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:22:27,192 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130456.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:22:38,657 INFO [optim.py:369] (0/4) 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 19:22:44,115 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130477.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:22:45,333 INFO [train.py:901] (0/4) Epoch 17, batch 1150, loss[loss=0.2457, simple_loss=0.3105, pruned_loss=0.09046, over 7330.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2968, pruned_loss=0.06825, over 1616850.70 frames. ], batch size: 72, lr: 4.52e-03, grad_scale: 8.0 +2023-02-06 19:22:45,507 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130479.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:22:45,954 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-06 19:23:16,161 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130524.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:23:19,435 INFO [train.py:901] (0/4) Epoch 17, batch 1200, loss[loss=0.1955, simple_loss=0.2663, pruned_loss=0.06232, over 7261.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.2959, pruned_loss=0.06829, over 1613428.08 frames. ], batch size: 16, lr: 4.52e-03, grad_scale: 8.0 +2023-02-06 19:23:33,403 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130549.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:23:45,141 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130566.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:23:46,148 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 +2023-02-06 19:23:47,784 INFO [optim.py:369] (0/4) 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 19:23:53,876 INFO [train.py:901] (0/4) Epoch 17, batch 1250, loss[loss=0.2131, simple_loss=0.2824, pruned_loss=0.07189, over 7559.00 frames. ], tot_loss[loss=0.2173, simple_loss=0.2971, pruned_loss=0.06878, over 1615166.50 frames. ], batch size: 18, lr: 4.52e-03, grad_scale: 8.0 +2023-02-06 19:23:57,434 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130583.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:24:30,844 INFO [train.py:901] (0/4) Epoch 17, batch 1300, loss[loss=0.2392, simple_loss=0.314, pruned_loss=0.08224, over 8332.00 frames. ], tot_loss[loss=0.2175, simple_loss=0.2972, pruned_loss=0.06884, over 1616693.59 frames. ], batch size: 25, lr: 4.52e-03, grad_scale: 8.0 +2023-02-06 19:24:59,327 INFO [optim.py:369] (0/4) 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 19:25:05,684 INFO [train.py:901] (0/4) Epoch 17, batch 1350, loss[loss=0.2258, simple_loss=0.2862, pruned_loss=0.08273, over 7454.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2957, pruned_loss=0.06796, over 1614663.38 frames. ], batch size: 17, lr: 4.52e-03, grad_scale: 8.0 +2023-02-06 19:25:27,719 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3481, 1.3055, 2.3761, 1.2037, 2.3592, 2.5350, 2.6726, 2.1719], + device='cuda:0'), covar=tensor([0.1011, 0.1232, 0.0422, 0.1889, 0.0618, 0.0364, 0.0656, 0.0689], + device='cuda:0'), in_proj_covar=tensor([0.0284, 0.0310, 0.0275, 0.0304, 0.0297, 0.0254, 0.0390, 0.0299], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-06 19:25:43,058 INFO [train.py:901] (0/4) Epoch 17, batch 1400, loss[loss=0.1986, simple_loss=0.2773, pruned_loss=0.05992, over 7778.00 frames. ], tot_loss[loss=0.2161, simple_loss=0.2961, pruned_loss=0.06808, over 1617269.74 frames. ], batch size: 19, lr: 4.52e-03, grad_scale: 8.0 +2023-02-06 19:25:45,981 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130733.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:25:47,357 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130735.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:25:54,821 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=130746.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:26:03,085 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130758.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:26:04,417 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130760.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:26:11,022 INFO [optim.py:369] (0/4) 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 19:26:17,376 INFO [train.py:901] (0/4) Epoch 17, batch 1450, loss[loss=0.2844, simple_loss=0.3431, pruned_loss=0.1129, over 7171.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.2958, pruned_loss=0.06829, over 1612295.90 frames. ], batch size: 71, lr: 4.52e-03, grad_scale: 8.0 +2023-02-06 19:26:20,731 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-06 19:26:27,816 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=130794.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:26:32,211 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=130800.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:26:54,259 INFO [train.py:901] (0/4) Epoch 17, batch 1500, loss[loss=0.275, simple_loss=0.3392, pruned_loss=0.1054, over 8575.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.2949, pruned_loss=0.06777, over 1617452.16 frames. ], batch size: 34, lr: 4.52e-03, grad_scale: 8.0 +2023-02-06 19:27:17,117 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130861.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:27:22,933 INFO [optim.py:369] (0/4) 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 19:27:29,112 INFO [train.py:901] (0/4) Epoch 17, batch 1550, loss[loss=0.2481, simple_loss=0.3249, pruned_loss=0.08566, over 8333.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.2954, pruned_loss=0.06754, over 1620814.19 frames. ], batch size: 25, lr: 4.52e-03, grad_scale: 8.0 +2023-02-06 19:27:50,125 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130909.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:27:50,698 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=130910.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:27:54,298 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130915.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:28:02,609 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=130927.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:28:03,803 INFO [train.py:901] (0/4) Epoch 17, batch 1600, loss[loss=0.2149, simple_loss=0.2961, pruned_loss=0.06687, over 8235.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.2956, pruned_loss=0.06842, over 1612604.75 frames. ], batch size: 22, lr: 4.52e-03, grad_scale: 8.0 +2023-02-06 19:28:34,756 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 17, batch 1650, loss[loss=0.2311, simple_loss=0.3101, pruned_loss=0.076, over 8091.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2961, pruned_loss=0.06833, over 1616934.81 frames. ], batch size: 21, lr: 4.51e-03, grad_scale: 16.0 +2023-02-06 19:28:57,393 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 +2023-02-06 19:29:13,494 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131025.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:29:16,141 INFO [train.py:901] (0/4) Epoch 17, batch 1700, loss[loss=0.1891, simple_loss=0.268, pruned_loss=0.05507, over 7698.00 frames. ], tot_loss[loss=0.2154, simple_loss=0.2953, pruned_loss=0.06777, over 1614112.68 frames. ], batch size: 18, lr: 4.51e-03, grad_scale: 16.0 +2023-02-06 19:29:25,393 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131042.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:29:46,942 INFO [optim.py:369] (0/4) 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 19:29:53,061 INFO [train.py:901] (0/4) Epoch 17, batch 1750, loss[loss=0.2335, simple_loss=0.3119, pruned_loss=0.07757, over 8511.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.2962, pruned_loss=0.06823, over 1619160.17 frames. ], batch size: 28, lr: 4.51e-03, grad_scale: 16.0 +2023-02-06 19:29:58,715 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0746, 1.5527, 1.6739, 1.4547, 0.9426, 1.5260, 1.8248, 1.4775], + device='cuda:0'), covar=tensor([0.0550, 0.1179, 0.1706, 0.1414, 0.0636, 0.1464, 0.0684, 0.0660], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0152, 0.0191, 0.0157, 0.0100, 0.0164, 0.0115, 0.0140], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 19:30:19,609 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131117.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:30:27,902 INFO [train.py:901] (0/4) Epoch 17, batch 1800, loss[loss=0.1916, simple_loss=0.2832, pruned_loss=0.05003, over 8244.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.296, pruned_loss=0.06827, over 1617982.89 frames. ], batch size: 24, lr: 4.51e-03, grad_scale: 16.0 +2023-02-06 19:30:37,097 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131142.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:30:51,223 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.4050, 1.3301, 3.7724, 1.5723, 2.9609, 3.0089, 3.4256, 3.3488], + device='cuda:0'), covar=tensor([0.1539, 0.6277, 0.1398, 0.4866, 0.2399, 0.1744, 0.1168, 0.1296], + device='cuda:0'), in_proj_covar=tensor([0.0568, 0.0622, 0.0654, 0.0592, 0.0671, 0.0575, 0.0567, 0.0635], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 19:30:52,690 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131165.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:30:55,946 INFO [optim.py:369] (0/4) 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] (0/4) 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] (0/4) Epoch 17, batch 1850, loss[loss=0.2041, simple_loss=0.2718, pruned_loss=0.06824, over 7708.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.2967, pruned_loss=0.06842, over 1622886.14 frames. ], batch size: 18, lr: 4.51e-03, grad_scale: 16.0 +2023-02-06 19:31:12,468 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131190.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:31:13,829 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=131192.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:31:16,687 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131196.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:31:37,446 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7343, 1.8496, 2.4382, 1.5493, 1.2984, 2.3961, 0.4541, 1.4554], + device='cuda:0'), covar=tensor([0.2074, 0.1424, 0.0352, 0.1472, 0.3028, 0.0425, 0.2491, 0.1508], + device='cuda:0'), in_proj_covar=tensor([0.0179, 0.0183, 0.0115, 0.0216, 0.0261, 0.0122, 0.0166, 0.0179], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 19:31:39,958 INFO [train.py:901] (0/4) Epoch 17, batch 1900, loss[loss=0.2157, simple_loss=0.3089, pruned_loss=0.06121, over 8455.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.2973, pruned_loss=0.06879, over 1621610.57 frames. ], batch size: 25, lr: 4.51e-03, grad_scale: 16.0 +2023-02-06 19:32:08,105 INFO [optim.py:369] (0/4) 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 19:32:08,137 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-06 19:32:14,113 INFO [train.py:901] (0/4) Epoch 17, batch 1950, loss[loss=0.2013, simple_loss=0.2811, pruned_loss=0.06077, over 7979.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.2969, pruned_loss=0.06897, over 1616705.50 frames. ], batch size: 21, lr: 4.51e-03, grad_scale: 16.0 +2023-02-06 19:32:15,744 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131281.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:32:19,629 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-06 19:32:28,898 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131298.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:32:35,128 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131306.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:32:39,911 WARNING [train.py:1067] (0/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-06 19:32:47,544 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131323.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:32:51,199 INFO [train.py:901] (0/4) Epoch 17, batch 2000, loss[loss=0.2209, simple_loss=0.3157, pruned_loss=0.06307, over 7974.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.2961, pruned_loss=0.06826, over 1617210.81 frames. ], batch size: 21, lr: 4.51e-03, grad_scale: 16.0 +2023-02-06 19:33:19,856 INFO [optim.py:369] (0/4) 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 19:33:25,353 INFO [train.py:901] (0/4) Epoch 17, batch 2050, loss[loss=0.212, simple_loss=0.2955, pruned_loss=0.06425, over 8506.00 frames. ], tot_loss[loss=0.2154, simple_loss=0.2957, pruned_loss=0.06758, over 1620948.13 frames. ], batch size: 26, lr: 4.51e-03, grad_scale: 8.0 +2023-02-06 19:34:00,658 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=131427.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:34:01,953 INFO [train.py:901] (0/4) Epoch 17, batch 2100, loss[loss=0.2201, simple_loss=0.3099, pruned_loss=0.06517, over 8522.00 frames. ], tot_loss[loss=0.2159, simple_loss=0.2958, pruned_loss=0.06796, over 1618423.65 frames. ], batch size: 31, lr: 4.51e-03, grad_scale: 8.0 +2023-02-06 19:34:06,118 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=131434.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:34:31,418 INFO [optim.py:369] (0/4) 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 19:34:36,975 INFO [train.py:901] (0/4) Epoch 17, batch 2150, loss[loss=0.2182, simple_loss=0.301, pruned_loss=0.06767, over 8233.00 frames. ], tot_loss[loss=0.2165, simple_loss=0.2965, pruned_loss=0.06829, over 1619541.29 frames. ], batch size: 22, lr: 4.51e-03, grad_scale: 8.0 +2023-02-06 19:34:58,678 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=131510.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:35:12,360 INFO [train.py:901] (0/4) Epoch 17, batch 2200, loss[loss=0.1991, simple_loss=0.2792, pruned_loss=0.05951, over 7935.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2967, pruned_loss=0.0683, over 1617068.54 frames. ], batch size: 20, lr: 4.50e-03, grad_scale: 8.0 +2023-02-06 19:35:17,202 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=131536.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:35:36,119 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3547, 1.8141, 1.4089, 2.8674, 1.3828, 1.2656, 2.0711, 2.0669], + device='cuda:0'), covar=tensor([0.1622, 0.1369, 0.2113, 0.0416, 0.1252, 0.2053, 0.0889, 0.0921], + device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0198, 0.0247, 0.0211, 0.0208, 0.0246, 0.0253, 0.0209], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 19:35:43,528 INFO [optim.py:369] (0/4) 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 19:35:49,226 INFO [train.py:901] (0/4) Epoch 17, batch 2250, loss[loss=0.2273, simple_loss=0.3109, pruned_loss=0.07187, over 8462.00 frames. ], tot_loss[loss=0.217, simple_loss=0.2973, pruned_loss=0.06834, over 1617671.96 frames. ], batch size: 25, lr: 4.50e-03, grad_scale: 8.0 +2023-02-06 19:36:23,874 INFO [train.py:901] (0/4) Epoch 17, batch 2300, loss[loss=0.1887, simple_loss=0.2742, pruned_loss=0.05155, over 8085.00 frames. ], tot_loss[loss=0.2169, simple_loss=0.2973, pruned_loss=0.0683, over 1614991.99 frames. ], batch size: 21, lr: 4.50e-03, grad_scale: 8.0 +2023-02-06 19:36:24,059 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8852, 2.2555, 1.8887, 2.8422, 1.3284, 1.5852, 1.9190, 2.3511], + device='cuda:0'), covar=tensor([0.0844, 0.0777, 0.0937, 0.0386, 0.1107, 0.1400, 0.1009, 0.0690], + device='cuda:0'), in_proj_covar=tensor([0.0230, 0.0199, 0.0249, 0.0212, 0.0209, 0.0247, 0.0254, 0.0210], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 19:36:40,749 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131651.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:36:55,874 INFO [optim.py:369] (0/4) 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 19:37:01,535 INFO [train.py:901] (0/4) Epoch 17, batch 2350, loss[loss=0.2429, simple_loss=0.3227, pruned_loss=0.08162, over 8514.00 frames. ], tot_loss[loss=0.2175, simple_loss=0.2977, pruned_loss=0.06861, over 1617482.22 frames. ], batch size: 26, lr: 4.50e-03, grad_scale: 8.0 +2023-02-06 19:37:16,027 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7487, 1.7541, 2.3432, 1.6853, 1.3153, 2.2900, 0.7398, 1.5176], + device='cuda:0'), covar=tensor([0.1993, 0.1237, 0.0347, 0.1248, 0.2828, 0.0399, 0.2006, 0.1462], + device='cuda:0'), in_proj_covar=tensor([0.0179, 0.0183, 0.0114, 0.0216, 0.0260, 0.0121, 0.0166, 0.0179], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 19:37:21,847 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 +2023-02-06 19:37:35,922 INFO [train.py:901] (0/4) Epoch 17, batch 2400, loss[loss=0.2219, simple_loss=0.3091, pruned_loss=0.06738, over 8543.00 frames. ], tot_loss[loss=0.216, simple_loss=0.2957, pruned_loss=0.06814, over 1609576.63 frames. ], batch size: 49, lr: 4.50e-03, grad_scale: 8.0 +2023-02-06 19:37:40,381 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3422, 2.0592, 1.5787, 1.8512, 1.7573, 1.4473, 1.6820, 1.6803], + device='cuda:0'), covar=tensor([0.1023, 0.0414, 0.1094, 0.0499, 0.0569, 0.1232, 0.0748, 0.0780], + device='cuda:0'), in_proj_covar=tensor([0.0349, 0.0233, 0.0324, 0.0299, 0.0297, 0.0328, 0.0340, 0.0312], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 19:38:06,368 INFO [optim.py:369] (0/4) 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 19:38:06,488 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=131771.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:38:12,194 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=131778.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:38:12,838 INFO [train.py:901] (0/4) Epoch 17, batch 2450, loss[loss=0.1734, simple_loss=0.2608, pruned_loss=0.04296, over 7922.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2938, pruned_loss=0.06644, over 1611838.54 frames. ], batch size: 20, lr: 4.50e-03, grad_scale: 8.0 +2023-02-06 19:38:47,887 INFO [train.py:901] (0/4) Epoch 17, batch 2500, loss[loss=0.2195, simple_loss=0.3046, pruned_loss=0.06724, over 8285.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2927, pruned_loss=0.06544, over 1614836.91 frames. ], batch size: 23, lr: 4.50e-03, grad_scale: 8.0 +2023-02-06 19:38:58,558 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1350, 1.4807, 1.8006, 1.3687, 0.9269, 1.4943, 1.9072, 1.8781], + device='cuda:0'), covar=tensor([0.0435, 0.1210, 0.1504, 0.1372, 0.0581, 0.1429, 0.0646, 0.0541], + device='cuda:0'), in_proj_covar=tensor([0.0096, 0.0151, 0.0189, 0.0156, 0.0099, 0.0161, 0.0114, 0.0138], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 19:39:05,482 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 19:39:20,112 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 17, batch 2550, loss[loss=0.1729, simple_loss=0.2448, pruned_loss=0.05044, over 7235.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2936, pruned_loss=0.06652, over 1614241.38 frames. ], batch size: 16, lr: 4.50e-03, grad_scale: 8.0 +2023-02-06 19:39:29,613 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131886.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:39:34,584 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131893.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:39:42,791 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8395, 1.7058, 3.3002, 1.5101, 2.5266, 3.5750, 3.6747, 2.9827], + device='cuda:0'), covar=tensor([0.1227, 0.1648, 0.0433, 0.2077, 0.0990, 0.0298, 0.0554, 0.0655], + device='cuda:0'), in_proj_covar=tensor([0.0282, 0.0308, 0.0275, 0.0302, 0.0292, 0.0252, 0.0388, 0.0299], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-06 19:39:45,565 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131907.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:40:00,840 INFO [train.py:901] (0/4) Epoch 17, batch 2600, loss[loss=0.2123, simple_loss=0.3024, pruned_loss=0.06111, over 8286.00 frames. ], tot_loss[loss=0.2143, simple_loss=0.2944, pruned_loss=0.06716, over 1611800.21 frames. ], batch size: 23, lr: 4.50e-03, grad_scale: 8.0 +2023-02-06 19:40:03,013 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131932.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:40:28,777 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131969.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:40:29,952 INFO [optim.py:369] (0/4) 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 19:40:35,439 INFO [train.py:901] (0/4) Epoch 17, batch 2650, loss[loss=0.2094, simple_loss=0.2796, pruned_loss=0.06963, over 7551.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2951, pruned_loss=0.06761, over 1611966.34 frames. ], batch size: 18, lr: 4.50e-03, grad_scale: 8.0 +2023-02-06 19:40:49,773 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-132000.pt +2023-02-06 19:40:52,326 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([0.9716, 1.2203, 1.2162, 0.5939, 1.2047, 0.9960, 0.0478, 1.1528], + device='cuda:0'), covar=tensor([0.0373, 0.0362, 0.0269, 0.0483, 0.0373, 0.0783, 0.0719, 0.0282], + device='cuda:0'), in_proj_covar=tensor([0.0426, 0.0368, 0.0315, 0.0425, 0.0351, 0.0509, 0.0376, 0.0391], + device='cuda:0'), out_proj_covar=tensor([1.1651e-04, 9.7932e-05, 8.3442e-05, 1.1367e-04, 9.4395e-05, 1.4685e-04, + 1.0248e-04, 1.0514e-04], device='cuda:0') +2023-02-06 19:41:13,509 INFO [train.py:901] (0/4) Epoch 17, batch 2700, loss[loss=0.2189, simple_loss=0.2923, pruned_loss=0.07272, over 7936.00 frames. ], tot_loss[loss=0.2166, simple_loss=0.2966, pruned_loss=0.06833, over 1616255.45 frames. ], batch size: 20, lr: 4.50e-03, grad_scale: 8.0 +2023-02-06 19:41:27,356 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=132049.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:41:31,121 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-02-06 19:41:42,493 INFO [optim.py:369] (0/4) 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 19:41:43,388 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=132072.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:41:48,257 INFO [train.py:901] (0/4) Epoch 17, batch 2750, loss[loss=0.2355, simple_loss=0.3038, pruned_loss=0.08363, over 7818.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.2959, pruned_loss=0.06824, over 1615512.66 frames. ], batch size: 20, lr: 4.50e-03, grad_scale: 8.0 +2023-02-06 19:42:25,035 INFO [train.py:901] (0/4) Epoch 17, batch 2800, loss[loss=0.2608, simple_loss=0.3329, pruned_loss=0.09433, over 6937.00 frames. ], tot_loss[loss=0.2161, simple_loss=0.2953, pruned_loss=0.06844, over 1611853.98 frames. ], batch size: 71, lr: 4.49e-03, grad_scale: 8.0 +2023-02-06 19:42:35,383 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132142.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:42:40,200 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132149.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:42:52,677 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132167.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:42:55,269 INFO [optim.py:369] (0/4) 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 19:42:57,335 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132174.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:43:00,621 INFO [train.py:901] (0/4) Epoch 17, batch 2850, loss[loss=0.1952, simple_loss=0.2865, pruned_loss=0.05192, over 8561.00 frames. ], tot_loss[loss=0.2171, simple_loss=0.2963, pruned_loss=0.069, over 1613652.59 frames. ], batch size: 31, lr: 4.49e-03, grad_scale: 8.0 +2023-02-06 19:43:06,963 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.67 vs. limit=5.0 +2023-02-06 19:43:22,885 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.5378, 1.8915, 4.4181, 1.9715, 2.5652, 5.0209, 5.0140, 4.2847], + device='cuda:0'), covar=tensor([0.1080, 0.1745, 0.0295, 0.2026, 0.1129, 0.0171, 0.0449, 0.0543], + device='cuda:0'), in_proj_covar=tensor([0.0283, 0.0309, 0.0276, 0.0303, 0.0293, 0.0253, 0.0388, 0.0299], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-06 19:43:29,199 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=132219.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:43:33,399 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132225.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:43:35,744 INFO [train.py:901] (0/4) Epoch 17, batch 2900, loss[loss=0.2416, simple_loss=0.3257, pruned_loss=0.07877, over 8474.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.296, pruned_loss=0.06871, over 1612873.21 frames. ], batch size: 25, lr: 4.49e-03, grad_scale: 8.0 +2023-02-06 19:43:52,922 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132250.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:44:08,362 INFO [optim.py:369] (0/4) 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 19:44:09,811 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-06 19:44:13,735 INFO [train.py:901] (0/4) Epoch 17, batch 2950, loss[loss=0.2033, simple_loss=0.2782, pruned_loss=0.06421, over 7933.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.2955, pruned_loss=0.06798, over 1615243.36 frames. ], batch size: 20, lr: 4.49e-03, grad_scale: 8.0 +2023-02-06 19:44:42,851 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0340, 1.7235, 2.6460, 1.6215, 2.2656, 2.8933, 2.8632, 2.5992], + device='cuda:0'), covar=tensor([0.0847, 0.1366, 0.0709, 0.1678, 0.1371, 0.0256, 0.0665, 0.0474], + device='cuda:0'), in_proj_covar=tensor([0.0283, 0.0309, 0.0275, 0.0302, 0.0292, 0.0252, 0.0388, 0.0298], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-06 19:44:48,301 INFO [train.py:901] (0/4) Epoch 17, batch 3000, loss[loss=0.2117, simple_loss=0.2921, pruned_loss=0.06569, over 8554.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2961, pruned_loss=0.06863, over 1613103.95 frames. ], batch size: 31, lr: 4.49e-03, grad_scale: 8.0 +2023-02-06 19:44:48,302 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 19:45:00,592 INFO [train.py:935] (0/4) Epoch 17, validation: loss=0.1786, simple_loss=0.2786, pruned_loss=0.03928, over 944034.00 frames. +2023-02-06 19:45:00,593 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-06 19:45:04,437 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132334.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:45:31,437 INFO [optim.py:369] (0/4) 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 19:45:37,096 INFO [train.py:901] (0/4) Epoch 17, batch 3050, loss[loss=0.2233, simple_loss=0.303, pruned_loss=0.07177, over 8448.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2961, pruned_loss=0.06862, over 1613590.28 frames. ], batch size: 27, lr: 4.49e-03, grad_scale: 8.0 +2023-02-06 19:45:48,262 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=132416.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:46:12,935 INFO [train.py:901] (0/4) Epoch 17, batch 3100, loss[loss=0.2199, simple_loss=0.293, pruned_loss=0.07336, over 7442.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.2971, pruned_loss=0.06882, over 1613943.02 frames. ], batch size: 17, lr: 4.49e-03, grad_scale: 8.0 +2023-02-06 19:46:41,889 INFO [optim.py:369] (0/4) 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 19:46:47,327 INFO [train.py:901] (0/4) Epoch 17, batch 3150, loss[loss=0.2115, simple_loss=0.2987, pruned_loss=0.06218, over 8290.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.2979, pruned_loss=0.06911, over 1619583.78 frames. ], batch size: 23, lr: 4.49e-03, grad_scale: 8.0 +2023-02-06 19:47:05,433 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6551, 1.8319, 1.6084, 2.2695, 1.0174, 1.3866, 1.5966, 1.8523], + device='cuda:0'), covar=tensor([0.0813, 0.0810, 0.0947, 0.0434, 0.1209, 0.1444, 0.0849, 0.0731], + device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0200, 0.0249, 0.0212, 0.0211, 0.0248, 0.0254, 0.0211], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 19:47:09,721 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132508.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:47:24,997 INFO [train.py:901] (0/4) Epoch 17, batch 3200, loss[loss=0.2099, simple_loss=0.2776, pruned_loss=0.07108, over 7791.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.297, pruned_loss=0.06833, over 1618978.80 frames. ], batch size: 19, lr: 4.49e-03, grad_scale: 8.0 +2023-02-06 19:47:26,582 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132531.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:47:54,177 INFO [optim.py:369] (0/4) 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 19:47:59,504 INFO [train.py:901] (0/4) Epoch 17, batch 3250, loss[loss=0.2064, simple_loss=0.2799, pruned_loss=0.06647, over 7795.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.2961, pruned_loss=0.06825, over 1612361.73 frames. ], batch size: 19, lr: 4.49e-03, grad_scale: 8.0 +2023-02-06 19:48:07,398 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132590.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:48:26,241 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132615.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:48:37,505 INFO [train.py:901] (0/4) Epoch 17, batch 3300, loss[loss=0.1957, simple_loss=0.2828, pruned_loss=0.05435, over 8654.00 frames. ], tot_loss[loss=0.2165, simple_loss=0.296, pruned_loss=0.06846, over 1614468.47 frames. ], batch size: 34, lr: 4.49e-03, grad_scale: 8.0 +2023-02-06 19:48:58,465 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4246, 1.9755, 4.4509, 2.1481, 2.3657, 4.9321, 5.0555, 4.2734], + device='cuda:0'), covar=tensor([0.1096, 0.1539, 0.0266, 0.1955, 0.1206, 0.0185, 0.0427, 0.0492], + device='cuda:0'), in_proj_covar=tensor([0.0285, 0.0310, 0.0276, 0.0305, 0.0295, 0.0254, 0.0391, 0.0301], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 19:49:06,785 INFO [optim.py:369] (0/4) 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 19:49:07,214 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 +2023-02-06 19:49:11,856 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=132678.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:49:12,441 INFO [train.py:901] (0/4) Epoch 17, batch 3350, loss[loss=0.1619, simple_loss=0.2336, pruned_loss=0.04509, over 6845.00 frames. ], tot_loss[loss=0.2154, simple_loss=0.295, pruned_loss=0.06786, over 1611918.60 frames. ], batch size: 15, lr: 4.49e-03, grad_scale: 8.0 +2023-02-06 19:49:19,014 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.33 vs. limit=5.0 +2023-02-06 19:49:26,537 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3853, 2.6826, 1.8200, 2.3461, 2.2853, 1.6320, 2.1271, 2.2881], + device='cuda:0'), covar=tensor([0.1659, 0.0359, 0.1260, 0.0699, 0.0682, 0.1488, 0.0970, 0.1036], + device='cuda:0'), in_proj_covar=tensor([0.0349, 0.0231, 0.0321, 0.0298, 0.0295, 0.0326, 0.0338, 0.0312], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 19:49:49,260 INFO [train.py:901] (0/4) Epoch 17, batch 3400, loss[loss=0.205, simple_loss=0.2757, pruned_loss=0.06718, over 7817.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.2944, pruned_loss=0.06729, over 1611566.93 frames. ], batch size: 20, lr: 4.48e-03, grad_scale: 8.0 +2023-02-06 19:50:02,395 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 +2023-02-06 19:50:14,680 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=132764.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:50:14,758 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132764.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:50:19,442 INFO [optim.py:369] (0/4) 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 19:50:20,634 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 +2023-02-06 19:50:24,936 INFO [train.py:901] (0/4) Epoch 17, batch 3450, loss[loss=0.2341, simple_loss=0.3038, pruned_loss=0.08222, over 8464.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.295, pruned_loss=0.06804, over 1608051.52 frames. ], batch size: 25, lr: 4.48e-03, grad_scale: 8.0 +2023-02-06 19:50:30,854 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132787.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:50:32,181 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132789.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:50:47,288 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 +2023-02-06 19:50:47,670 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132812.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:50:59,811 INFO [train.py:901] (0/4) Epoch 17, batch 3500, loss[loss=0.2194, simple_loss=0.2947, pruned_loss=0.07202, over 8328.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.2946, pruned_loss=0.06788, over 1607220.35 frames. ], batch size: 25, lr: 4.48e-03, grad_scale: 8.0 +2023-02-06 19:51:13,841 WARNING [train.py:1067] (0/4) 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] (0/4) 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 19:51:37,025 INFO [train.py:901] (0/4) Epoch 17, batch 3550, loss[loss=0.2135, simple_loss=0.2814, pruned_loss=0.07277, over 7974.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.2951, pruned_loss=0.06805, over 1608155.03 frames. ], batch size: 21, lr: 4.48e-03, grad_scale: 8.0 +2023-02-06 19:51:54,233 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2292, 1.3058, 1.5398, 1.2432, 0.6948, 1.2785, 1.1639, 1.1007], + device='cuda:0'), covar=tensor([0.0602, 0.1298, 0.1759, 0.1453, 0.0615, 0.1612, 0.0725, 0.0691], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0153, 0.0191, 0.0157, 0.0100, 0.0163, 0.0115, 0.0139], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 19:52:11,140 INFO [train.py:901] (0/4) Epoch 17, batch 3600, loss[loss=0.2044, simple_loss=0.2873, pruned_loss=0.06072, over 7804.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.295, pruned_loss=0.06766, over 1612379.27 frames. ], batch size: 20, lr: 4.48e-03, grad_scale: 8.0 +2023-02-06 19:52:41,872 INFO [optim.py:369] (0/4) 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 19:52:48,333 INFO [train.py:901] (0/4) Epoch 17, batch 3650, loss[loss=0.254, simple_loss=0.3409, pruned_loss=0.0835, over 8501.00 frames. ], tot_loss[loss=0.2143, simple_loss=0.2942, pruned_loss=0.06714, over 1611875.70 frames. ], batch size: 26, lr: 4.48e-03, grad_scale: 8.0 +2023-02-06 19:53:17,248 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0568, 2.1894, 2.3325, 1.5421, 2.3844, 1.6206, 1.7591, 1.9662], + device='cuda:0'), covar=tensor([0.0542, 0.0320, 0.0188, 0.0509, 0.0361, 0.0644, 0.0565, 0.0345], + device='cuda:0'), in_proj_covar=tensor([0.0432, 0.0374, 0.0318, 0.0430, 0.0357, 0.0515, 0.0379, 0.0393], + device='cuda:0'), out_proj_covar=tensor([1.1798e-04, 9.9506e-05, 8.4217e-05, 1.1506e-04, 9.5856e-05, 1.4830e-04, + 1.0340e-04, 1.0538e-04], device='cuda:0') +2023-02-06 19:53:18,535 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=133022.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:53:21,766 WARNING [train.py:1067] (0/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-06 19:53:23,088 INFO [train.py:901] (0/4) Epoch 17, batch 3700, loss[loss=0.2115, simple_loss=0.2673, pruned_loss=0.07787, over 7439.00 frames. ], tot_loss[loss=0.214, simple_loss=0.294, pruned_loss=0.06697, over 1609546.71 frames. ], batch size: 17, lr: 4.48e-03, grad_scale: 8.0 +2023-02-06 19:53:53,577 INFO [optim.py:369] (0/4) 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 19:53:59,125 INFO [train.py:901] (0/4) Epoch 17, batch 3750, loss[loss=0.2331, simple_loss=0.3159, pruned_loss=0.07517, over 8743.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.2934, pruned_loss=0.06648, over 1615572.08 frames. ], batch size: 39, lr: 4.48e-03, grad_scale: 8.0 +2023-02-06 19:54:16,848 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3228, 1.5541, 2.1139, 1.2542, 1.3007, 1.6286, 1.4074, 1.3545], + device='cuda:0'), covar=tensor([0.1885, 0.2243, 0.0883, 0.4111, 0.1992, 0.3049, 0.2176, 0.2184], + device='cuda:0'), in_proj_covar=tensor([0.0506, 0.0565, 0.0543, 0.0613, 0.0631, 0.0568, 0.0508, 0.0623], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 19:54:21,509 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=133108.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:54:35,519 INFO [train.py:901] (0/4) Epoch 17, batch 3800, loss[loss=0.2287, simple_loss=0.3102, pruned_loss=0.07364, over 8426.00 frames. ], tot_loss[loss=0.2139, simple_loss=0.294, pruned_loss=0.06691, over 1614428.18 frames. ], batch size: 27, lr: 4.48e-03, grad_scale: 8.0 +2023-02-06 19:54:41,262 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=133137.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:55:04,572 INFO [optim.py:369] (0/4) 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 19:55:09,939 INFO [train.py:901] (0/4) Epoch 17, batch 3850, loss[loss=0.2838, simple_loss=0.3559, pruned_loss=0.1059, over 8245.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2956, pruned_loss=0.06801, over 1610979.48 frames. ], batch size: 24, lr: 4.48e-03, grad_scale: 8.0 +2023-02-06 19:55:31,145 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-06 19:55:42,996 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=133223.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:55:46,898 INFO [train.py:901] (0/4) Epoch 17, batch 3900, loss[loss=0.2248, simple_loss=0.3104, pruned_loss=0.06957, over 8527.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.2959, pruned_loss=0.0683, over 1609390.00 frames. ], batch size: 31, lr: 4.48e-03, grad_scale: 8.0 +2023-02-06 19:56:15,758 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 17, batch 3950, loss[loss=0.2348, simple_loss=0.2917, pruned_loss=0.08894, over 6798.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2951, pruned_loss=0.06826, over 1606243.92 frames. ], batch size: 15, lr: 4.48e-03, grad_scale: 8.0 +2023-02-06 19:56:56,940 INFO [train.py:901] (0/4) Epoch 17, batch 4000, loss[loss=0.2131, simple_loss=0.2845, pruned_loss=0.07081, over 7433.00 frames. ], tot_loss[loss=0.215, simple_loss=0.2943, pruned_loss=0.06783, over 1603954.47 frames. ], batch size: 17, lr: 4.47e-03, grad_scale: 8.0 +2023-02-06 19:57:27,418 INFO [optim.py:369] (0/4) 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 19:57:32,340 INFO [train.py:901] (0/4) Epoch 17, batch 4050, loss[loss=0.2493, simple_loss=0.3149, pruned_loss=0.09187, over 7009.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2954, pruned_loss=0.0681, over 1605670.13 frames. ], batch size: 71, lr: 4.47e-03, grad_scale: 8.0 +2023-02-06 19:57:41,372 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133392.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 19:57:42,099 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=133393.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:57:42,756 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133394.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:57:59,643 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=133418.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:58:07,774 INFO [train.py:901] (0/4) Epoch 17, batch 4100, loss[loss=0.1937, simple_loss=0.2817, pruned_loss=0.05285, over 8300.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.2962, pruned_loss=0.0681, over 1609017.64 frames. ], batch size: 23, lr: 4.47e-03, grad_scale: 8.0 +2023-02-06 19:58:40,085 INFO [optim.py:369] (0/4) 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 19:58:45,050 INFO [train.py:901] (0/4) Epoch 17, batch 4150, loss[loss=0.2524, simple_loss=0.3395, pruned_loss=0.08267, over 8185.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2965, pruned_loss=0.06816, over 1610003.28 frames. ], batch size: 23, lr: 4.47e-03, grad_scale: 8.0 +2023-02-06 19:58:45,258 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=133479.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:58:54,026 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133492.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:59:02,274 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=133504.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 19:59:19,513 INFO [train.py:901] (0/4) Epoch 17, batch 4200, loss[loss=0.2167, simple_loss=0.3009, pruned_loss=0.06627, over 8754.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.2962, pruned_loss=0.06808, over 1610144.81 frames. ], batch size: 30, lr: 4.47e-03, grad_scale: 8.0 +2023-02-06 19:59:20,466 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.59 vs. limit=2.0 +2023-02-06 19:59:32,477 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-06 19:59:51,075 INFO [optim.py:369] (0/4) 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 19:59:56,775 INFO [train.py:901] (0/4) Epoch 17, batch 4250, loss[loss=0.1659, simple_loss=0.2473, pruned_loss=0.04224, over 7664.00 frames. ], tot_loss[loss=0.2177, simple_loss=0.2976, pruned_loss=0.06888, over 1612083.02 frames. ], batch size: 19, lr: 4.47e-03, grad_scale: 8.0 +2023-02-06 19:59:57,447 WARNING [train.py:1067] (0/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-06 20:00:30,996 INFO [train.py:901] (0/4) Epoch 17, batch 4300, loss[loss=0.2017, simple_loss=0.2922, pruned_loss=0.05557, over 8478.00 frames. ], tot_loss[loss=0.2166, simple_loss=0.2967, pruned_loss=0.06822, over 1610652.09 frames. ], batch size: 29, lr: 4.47e-03, grad_scale: 8.0 +2023-02-06 20:00:46,261 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9260, 2.3309, 1.9474, 2.8851, 1.4067, 1.5483, 2.0761, 2.3575], + device='cuda:0'), covar=tensor([0.0837, 0.0888, 0.0917, 0.0370, 0.1211, 0.1433, 0.0950, 0.0746], + device='cuda:0'), in_proj_covar=tensor([0.0230, 0.0199, 0.0248, 0.0211, 0.0208, 0.0245, 0.0254, 0.0211], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 20:00:58,160 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 +2023-02-06 20:00:59,905 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133670.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:01:01,921 INFO [optim.py:369] (0/4) 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 20:01:06,898 INFO [train.py:901] (0/4) Epoch 17, batch 4350, loss[loss=0.1729, simple_loss=0.256, pruned_loss=0.04494, over 7915.00 frames. ], tot_loss[loss=0.2173, simple_loss=0.297, pruned_loss=0.06883, over 1614045.29 frames. ], batch size: 20, lr: 4.47e-03, grad_scale: 8.0 +2023-02-06 20:01:11,421 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 +2023-02-06 20:01:31,286 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-06 20:01:43,132 INFO [train.py:901] (0/4) Epoch 17, batch 4400, loss[loss=0.2047, simple_loss=0.3063, pruned_loss=0.05159, over 8494.00 frames. ], tot_loss[loss=0.2172, simple_loss=0.2971, pruned_loss=0.06868, over 1619143.95 frames. ], batch size: 26, lr: 4.47e-03, grad_scale: 8.0 +2023-02-06 20:01:48,083 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=133736.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 20:01:49,407 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=133738.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:02:06,820 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2061, 1.0778, 1.2746, 1.0621, 0.9338, 1.3045, 0.0463, 0.9400], + device='cuda:0'), covar=tensor([0.1976, 0.1464, 0.0528, 0.0884, 0.2943, 0.0600, 0.2620, 0.1286], + device='cuda:0'), in_proj_covar=tensor([0.0180, 0.0186, 0.0116, 0.0219, 0.0264, 0.0123, 0.0169, 0.0181], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 20:02:12,877 INFO [optim.py:369] (0/4) 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 20:02:12,922 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 17, batch 4450, loss[loss=0.2316, simple_loss=0.3144, pruned_loss=0.07436, over 8100.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2962, pruned_loss=0.0683, over 1623279.94 frames. ], batch size: 23, lr: 4.47e-03, grad_scale: 8.0 +2023-02-06 20:02:55,023 INFO [train.py:901] (0/4) Epoch 17, batch 4500, loss[loss=0.2142, simple_loss=0.2857, pruned_loss=0.07133, over 7802.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.2962, pruned_loss=0.06806, over 1622723.51 frames. ], batch size: 20, lr: 4.47e-03, grad_scale: 8.0 +2023-02-06 20:03:00,051 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=133836.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:03:00,823 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.5639, 3.4389, 3.1681, 2.0815, 3.0783, 3.2006, 3.2605, 2.8840], + device='cuda:0'), covar=tensor([0.0913, 0.0722, 0.1032, 0.3918, 0.0920, 0.1053, 0.1224, 0.1036], + device='cuda:0'), in_proj_covar=tensor([0.0500, 0.0410, 0.0418, 0.0512, 0.0405, 0.0411, 0.0400, 0.0357], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 20:03:10,399 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=133851.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 20:03:10,909 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-06 20:03:11,730 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=133853.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:03:24,312 INFO [optim.py:369] (0/4) 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 20:03:29,194 INFO [train.py:901] (0/4) Epoch 17, batch 4550, loss[loss=0.2299, simple_loss=0.3089, pruned_loss=0.07549, over 8496.00 frames. ], tot_loss[loss=0.2161, simple_loss=0.2958, pruned_loss=0.06818, over 1617496.41 frames. ], batch size: 28, lr: 4.47e-03, grad_scale: 8.0 +2023-02-06 20:03:35,072 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1952, 3.9212, 2.6538, 3.1139, 3.0571, 2.4537, 3.0807, 3.3008], + device='cuda:0'), covar=tensor([0.1577, 0.0296, 0.0850, 0.0631, 0.0645, 0.1193, 0.0966, 0.0882], + device='cuda:0'), in_proj_covar=tensor([0.0352, 0.0232, 0.0324, 0.0299, 0.0296, 0.0330, 0.0339, 0.0313], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 20:03:47,065 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4780, 2.5605, 1.7743, 2.1720, 2.2068, 1.5648, 1.9599, 2.0997], + device='cuda:0'), covar=tensor([0.1475, 0.0345, 0.1149, 0.0612, 0.0643, 0.1470, 0.0910, 0.0913], + device='cuda:0'), in_proj_covar=tensor([0.0353, 0.0233, 0.0325, 0.0301, 0.0297, 0.0331, 0.0340, 0.0315], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 20:04:04,555 INFO [train.py:901] (0/4) Epoch 17, batch 4600, loss[loss=0.251, simple_loss=0.3101, pruned_loss=0.09601, over 7822.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.2959, pruned_loss=0.06826, over 1617697.07 frames. ], batch size: 20, lr: 4.46e-03, grad_scale: 8.0 +2023-02-06 20:04:21,366 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=133951.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:04:35,425 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 17, batch 4650, loss[loss=0.1977, simple_loss=0.2826, pruned_loss=0.05637, over 7928.00 frames. ], tot_loss[loss=0.2138, simple_loss=0.2938, pruned_loss=0.06693, over 1616212.73 frames. ], batch size: 20, lr: 4.46e-03, grad_scale: 8.0 +2023-02-06 20:04:46,113 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.24 vs. limit=5.0 +2023-02-06 20:04:54,999 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-134000.pt +2023-02-06 20:05:06,443 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=134014.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:05:07,186 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134015.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:05:16,655 INFO [train.py:901] (0/4) Epoch 17, batch 4700, loss[loss=0.244, simple_loss=0.3214, pruned_loss=0.0833, over 8500.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.295, pruned_loss=0.06811, over 1608223.50 frames. ], batch size: 39, lr: 4.46e-03, grad_scale: 8.0 +2023-02-06 20:05:24,863 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-02-06 20:05:28,194 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.6016, 2.3609, 3.3527, 2.6022, 3.1770, 2.5429, 2.1816, 1.9068], + device='cuda:0'), covar=tensor([0.4756, 0.4735, 0.1654, 0.3371, 0.2328, 0.2711, 0.1719, 0.5124], + device='cuda:0'), in_proj_covar=tensor([0.0921, 0.0936, 0.0774, 0.0904, 0.0967, 0.0853, 0.0725, 0.0805], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 20:05:37,702 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 +2023-02-06 20:05:48,989 INFO [optim.py:369] (0/4) 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 20:05:53,972 INFO [train.py:901] (0/4) Epoch 17, batch 4750, loss[loss=0.1768, simple_loss=0.2597, pruned_loss=0.04689, over 7784.00 frames. ], tot_loss[loss=0.2139, simple_loss=0.2935, pruned_loss=0.06712, over 1605682.35 frames. ], batch size: 19, lr: 4.46e-03, grad_scale: 8.0 +2023-02-06 20:06:13,289 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134107.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 20:06:14,640 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134109.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:06:17,892 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-06 20:06:20,569 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-06 20:06:20,742 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6974, 2.2430, 1.8840, 4.2608, 1.7157, 1.6371, 2.5437, 2.8328], + device='cuda:0'), covar=tensor([0.1743, 0.1389, 0.2108, 0.0212, 0.1428, 0.1988, 0.1074, 0.0942], + device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0199, 0.0250, 0.0211, 0.0209, 0.0247, 0.0254, 0.0212], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 20:06:28,172 INFO [train.py:901] (0/4) Epoch 17, batch 4800, loss[loss=0.2351, simple_loss=0.3185, pruned_loss=0.0758, over 8337.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2933, pruned_loss=0.06672, over 1608770.15 frames. ], batch size: 26, lr: 4.46e-03, grad_scale: 8.0 +2023-02-06 20:06:28,365 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134129.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:06:31,292 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=134132.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 20:06:32,668 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=134134.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:06:39,709 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1852, 2.7457, 3.1888, 1.7193, 3.3260, 1.9498, 1.5340, 2.0728], + device='cuda:0'), covar=tensor([0.0823, 0.0352, 0.0319, 0.0691, 0.0419, 0.0816, 0.0906, 0.0570], + device='cuda:0'), in_proj_covar=tensor([0.0433, 0.0372, 0.0316, 0.0427, 0.0354, 0.0512, 0.0377, 0.0396], + device='cuda:0'), out_proj_covar=tensor([1.1804e-04, 9.9210e-05, 8.3606e-05, 1.1416e-04, 9.4949e-05, 1.4767e-04, + 1.0283e-04, 1.0600e-04], device='cuda:0') +2023-02-06 20:07:00,736 INFO [optim.py:369] (0/4) 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 20:07:01,895 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.34 vs. limit=5.0 +2023-02-06 20:07:06,361 INFO [train.py:901] (0/4) Epoch 17, batch 4850, loss[loss=0.2012, simple_loss=0.2843, pruned_loss=0.05902, over 7928.00 frames. ], tot_loss[loss=0.2137, simple_loss=0.2932, pruned_loss=0.06714, over 1609070.42 frames. ], batch size: 20, lr: 4.46e-03, grad_scale: 8.0 +2023-02-06 20:07:14,640 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-06 20:07:26,226 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134207.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:07:41,126 INFO [train.py:901] (0/4) Epoch 17, batch 4900, loss[loss=0.234, simple_loss=0.3146, pruned_loss=0.07667, over 8080.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2937, pruned_loss=0.06736, over 1606993.97 frames. ], batch size: 21, lr: 4.46e-03, grad_scale: 8.0 +2023-02-06 20:07:43,465 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=134232.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:08:13,108 INFO [optim.py:369] (0/4) 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 20:08:17,811 INFO [train.py:901] (0/4) Epoch 17, batch 4950, loss[loss=0.2023, simple_loss=0.2845, pruned_loss=0.06001, over 8338.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.2942, pruned_loss=0.06751, over 1607407.51 frames. ], batch size: 26, lr: 4.46e-03, grad_scale: 8.0 +2023-02-06 20:08:36,229 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.8828, 2.6635, 3.6736, 2.0596, 1.9001, 3.6560, 0.7447, 2.0916], + device='cuda:0'), covar=tensor([0.1592, 0.1125, 0.0276, 0.1741, 0.2866, 0.0296, 0.2843, 0.1656], + device='cuda:0'), in_proj_covar=tensor([0.0177, 0.0184, 0.0116, 0.0218, 0.0262, 0.0123, 0.0167, 0.0181], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 20:08:49,532 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134322.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:08:52,975 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134327.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:08:54,216 INFO [train.py:901] (0/4) Epoch 17, batch 5000, loss[loss=0.2081, simple_loss=0.3017, pruned_loss=0.0573, over 8575.00 frames. ], tot_loss[loss=0.2139, simple_loss=0.2938, pruned_loss=0.06702, over 1606885.31 frames. ], batch size: 39, lr: 4.46e-03, grad_scale: 8.0 +2023-02-06 20:09:15,199 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=134359.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:09:24,851 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 17, batch 5050, loss[loss=0.2432, simple_loss=0.3075, pruned_loss=0.08947, over 7817.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.2942, pruned_loss=0.06742, over 1607969.78 frames. ], batch size: 20, lr: 4.46e-03, grad_scale: 8.0 +2023-02-06 20:09:35,086 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134385.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:09:54,046 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=134410.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:09:58,690 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-06 20:10:07,174 INFO [train.py:901] (0/4) Epoch 17, batch 5100, loss[loss=0.1867, simple_loss=0.2684, pruned_loss=0.05254, over 7917.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2937, pruned_loss=0.06737, over 1607504.27 frames. ], batch size: 20, lr: 4.46e-03, grad_scale: 8.0 +2023-02-06 20:10:36,993 INFO [optim.py:369] (0/4) 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 20:10:38,539 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134474.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:10:42,691 INFO [train.py:901] (0/4) Epoch 17, batch 5150, loss[loss=0.2927, simple_loss=0.3556, pruned_loss=0.1149, over 7235.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2934, pruned_loss=0.06677, over 1606367.59 frames. ], batch size: 71, lr: 4.46e-03, grad_scale: 8.0 +2023-02-06 20:11:11,937 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0800, 2.5615, 2.9068, 1.6809, 3.1820, 1.7560, 1.4423, 2.0311], + device='cuda:0'), covar=tensor([0.0750, 0.0371, 0.0260, 0.0697, 0.0465, 0.0886, 0.0830, 0.0528], + device='cuda:0'), in_proj_covar=tensor([0.0440, 0.0378, 0.0321, 0.0436, 0.0362, 0.0524, 0.0383, 0.0401], + device='cuda:0'), out_proj_covar=tensor([1.2011e-04, 1.0062e-04, 8.4996e-05, 1.1665e-04, 9.7002e-05, 1.5114e-04, + 1.0434e-04, 1.0736e-04], device='cuda:0') +2023-02-06 20:11:20,267 INFO [train.py:901] (0/4) Epoch 17, batch 5200, loss[loss=0.2198, simple_loss=0.3031, pruned_loss=0.06823, over 8773.00 frames. ], tot_loss[loss=0.2137, simple_loss=0.2937, pruned_loss=0.06687, over 1605547.16 frames. ], batch size: 30, lr: 4.45e-03, grad_scale: 8.0 +2023-02-06 20:11:37,249 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6057, 1.9717, 2.1455, 1.3320, 2.2697, 1.5232, 0.6449, 1.8496], + device='cuda:0'), covar=tensor([0.0564, 0.0291, 0.0224, 0.0499, 0.0321, 0.0771, 0.0723, 0.0251], + device='cuda:0'), in_proj_covar=tensor([0.0440, 0.0378, 0.0321, 0.0436, 0.0362, 0.0524, 0.0383, 0.0401], + device='cuda:0'), out_proj_covar=tensor([1.2000e-04, 1.0071e-04, 8.5080e-05, 1.1664e-04, 9.7036e-05, 1.5117e-04, + 1.0427e-04, 1.0740e-04], device='cuda:0') +2023-02-06 20:11:49,976 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 17, batch 5250, loss[loss=0.2401, simple_loss=0.3167, pruned_loss=0.08172, over 8349.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2954, pruned_loss=0.06813, over 1603755.40 frames. ], batch size: 26, lr: 4.45e-03, grad_scale: 8.0 +2023-02-06 20:11:57,582 WARNING [train.py:1067] (0/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-06 20:12:31,030 INFO [train.py:901] (0/4) Epoch 17, batch 5300, loss[loss=0.193, simple_loss=0.2794, pruned_loss=0.05332, over 7928.00 frames. ], tot_loss[loss=0.2153, simple_loss=0.2948, pruned_loss=0.06791, over 1602062.73 frames. ], batch size: 20, lr: 4.45e-03, grad_scale: 8.0 +2023-02-06 20:12:52,399 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.16 vs. limit=5.0 +2023-02-06 20:12:58,377 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=134666.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:13:01,784 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=134671.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:13:02,349 INFO [optim.py:369] (0/4) 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 20:13:07,143 INFO [train.py:901] (0/4) Epoch 17, batch 5350, loss[loss=0.2395, simple_loss=0.3069, pruned_loss=0.08605, over 8505.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2932, pruned_loss=0.06692, over 1601466.70 frames. ], batch size: 28, lr: 4.45e-03, grad_scale: 8.0 +2023-02-06 20:13:10,128 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.63 vs. limit=2.0 +2023-02-06 20:13:43,340 INFO [train.py:901] (0/4) Epoch 17, batch 5400, loss[loss=0.2123, simple_loss=0.3062, pruned_loss=0.05914, over 8195.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.2928, pruned_loss=0.0667, over 1602101.52 frames. ], batch size: 23, lr: 4.45e-03, grad_scale: 8.0 +2023-02-06 20:13:44,284 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134730.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:14:01,962 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=134755.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:14:14,310 INFO [optim.py:369] (0/4) 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 20:14:18,976 INFO [train.py:901] (0/4) Epoch 17, batch 5450, loss[loss=0.2836, simple_loss=0.3482, pruned_loss=0.1095, over 8359.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.2941, pruned_loss=0.06763, over 1605385.37 frames. ], batch size: 24, lr: 4.45e-03, grad_scale: 8.0 +2023-02-06 20:14:20,493 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134781.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:14:23,997 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134786.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:14:25,745 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 +2023-02-06 20:14:47,238 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 +2023-02-06 20:14:52,635 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.6840, 1.7402, 1.8298, 1.8352, 1.0022, 1.6386, 2.2642, 1.9855], + device='cuda:0'), covar=tensor([0.0445, 0.1274, 0.1721, 0.1295, 0.0642, 0.1539, 0.0655, 0.0590], + device='cuda:0'), in_proj_covar=tensor([0.0096, 0.0151, 0.0189, 0.0156, 0.0100, 0.0162, 0.0114, 0.0138], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 20:14:54,759 INFO [train.py:901] (0/4) Epoch 17, batch 5500, loss[loss=0.189, simple_loss=0.2654, pruned_loss=0.0563, over 7926.00 frames. ], tot_loss[loss=0.2141, simple_loss=0.2935, pruned_loss=0.06737, over 1607483.78 frames. ], batch size: 20, lr: 4.45e-03, grad_scale: 8.0 +2023-02-06 20:14:55,403 WARNING [train.py:1067] (0/4) 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] (0/4) 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] (0/4) Epoch 17, batch 5550, loss[loss=0.2273, simple_loss=0.3156, pruned_loss=0.06946, over 8456.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.294, pruned_loss=0.06747, over 1606578.77 frames. ], batch size: 27, lr: 4.45e-03, grad_scale: 8.0 +2023-02-06 20:15:32,206 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134880.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:15:42,711 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6881, 2.1227, 3.4132, 1.4998, 2.5549, 2.0649, 1.7604, 2.5049], + device='cuda:0'), covar=tensor([0.1742, 0.2283, 0.0732, 0.4042, 0.1739, 0.2904, 0.2030, 0.2163], + device='cuda:0'), in_proj_covar=tensor([0.0509, 0.0567, 0.0544, 0.0615, 0.0633, 0.0572, 0.0508, 0.0619], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 20:16:06,818 INFO [train.py:901] (0/4) Epoch 17, batch 5600, loss[loss=0.1775, simple_loss=0.2588, pruned_loss=0.04809, over 5973.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2924, pruned_loss=0.06664, over 1605027.23 frames. ], batch size: 13, lr: 4.45e-03, grad_scale: 8.0 +2023-02-06 20:16:38,756 INFO [optim.py:369] (0/4) 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 20:16:42,897 INFO [train.py:901] (0/4) Epoch 17, batch 5650, loss[loss=0.1939, simple_loss=0.2702, pruned_loss=0.05883, over 7538.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.2929, pruned_loss=0.06679, over 1607083.76 frames. ], batch size: 18, lr: 4.45e-03, grad_scale: 4.0 +2023-02-06 20:16:47,875 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.5725, 1.3698, 4.7375, 1.6624, 4.1915, 3.8895, 4.2951, 4.1446], + device='cuda:0'), covar=tensor([0.0497, 0.4768, 0.0491, 0.4059, 0.1145, 0.0857, 0.0546, 0.0650], + device='cuda:0'), in_proj_covar=tensor([0.0587, 0.0628, 0.0670, 0.0601, 0.0680, 0.0584, 0.0579, 0.0647], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 20:17:04,360 WARNING [train.py:1067] (0/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-06 20:17:18,727 INFO [train.py:901] (0/4) Epoch 17, batch 5700, loss[loss=0.2354, simple_loss=0.3092, pruned_loss=0.08086, over 8644.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2933, pruned_loss=0.0669, over 1608271.10 frames. ], batch size: 34, lr: 4.45e-03, grad_scale: 4.0 +2023-02-06 20:17:24,492 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=135037.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:17:28,004 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=135042.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:17:31,683 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.75 vs. limit=5.0 +2023-02-06 20:17:42,570 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=135062.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:17:45,910 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=135067.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:17:49,720 INFO [optim.py:369] (0/4) 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 20:17:53,692 INFO [train.py:901] (0/4) Epoch 17, batch 5750, loss[loss=0.2884, simple_loss=0.3585, pruned_loss=0.1091, over 8252.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.2928, pruned_loss=0.06667, over 1607856.25 frames. ], batch size: 24, lr: 4.45e-03, grad_scale: 4.0 +2023-02-06 20:18:11,538 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-06 20:18:14,099 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 +2023-02-06 20:18:30,180 INFO [train.py:901] (0/4) Epoch 17, batch 5800, loss[loss=0.2515, simple_loss=0.3301, pruned_loss=0.08647, over 8499.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2922, pruned_loss=0.0662, over 1605613.35 frames. ], batch size: 49, lr: 4.44e-03, grad_scale: 4.0 +2023-02-06 20:19:00,427 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.07 vs. limit=5.0 +2023-02-06 20:19:00,544 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 17, batch 5850, loss[loss=0.2683, simple_loss=0.3417, pruned_loss=0.09748, over 8033.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.2928, pruned_loss=0.06637, over 1608073.82 frames. ], batch size: 22, lr: 4.44e-03, grad_scale: 4.0 +2023-02-06 20:19:12,155 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=135189.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:19:37,618 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=135224.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:19:41,089 INFO [train.py:901] (0/4) Epoch 17, batch 5900, loss[loss=0.2307, simple_loss=0.3077, pruned_loss=0.07687, over 8193.00 frames. ], tot_loss[loss=0.2126, simple_loss=0.2923, pruned_loss=0.06649, over 1606131.00 frames. ], batch size: 23, lr: 4.44e-03, grad_scale: 4.0 +2023-02-06 20:20:12,430 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 17, batch 5950, loss[loss=0.2068, simple_loss=0.2789, pruned_loss=0.06733, over 7922.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2924, pruned_loss=0.06648, over 1610417.25 frames. ], batch size: 20, lr: 4.44e-03, grad_scale: 4.0 +2023-02-06 20:20:52,270 INFO [train.py:901] (0/4) Epoch 17, batch 6000, loss[loss=0.2128, simple_loss=0.292, pruned_loss=0.0668, over 7529.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2924, pruned_loss=0.0661, over 1606125.02 frames. ], batch size: 18, lr: 4.44e-03, grad_scale: 8.0 +2023-02-06 20:20:52,271 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 20:21:02,338 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.8493, 3.7661, 3.5271, 2.1765, 3.3932, 3.4824, 3.5180, 3.2812], + device='cuda:0'), covar=tensor([0.0950, 0.0589, 0.0889, 0.4653, 0.0965, 0.0905, 0.1160, 0.0915], + device='cuda:0'), in_proj_covar=tensor([0.0501, 0.0412, 0.0417, 0.0518, 0.0406, 0.0410, 0.0401, 0.0360], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 20:21:05,419 INFO [train.py:935] (0/4) Epoch 17, validation: loss=0.1774, simple_loss=0.2777, pruned_loss=0.03857, over 944034.00 frames. +2023-02-06 20:21:05,419 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-06 20:21:12,601 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=135339.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:21:36,686 INFO [optim.py:369] (0/4) 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 20:21:40,890 INFO [train.py:901] (0/4) Epoch 17, batch 6050, loss[loss=0.2277, simple_loss=0.3044, pruned_loss=0.07546, over 8501.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.293, pruned_loss=0.06644, over 1608155.32 frames. ], batch size: 28, lr: 4.44e-03, grad_scale: 8.0 +2023-02-06 20:21:41,163 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6570, 2.0176, 3.4362, 1.4681, 2.6799, 2.2449, 1.7219, 2.4403], + device='cuda:0'), covar=tensor([0.1843, 0.2539, 0.0865, 0.4427, 0.1733, 0.2786, 0.2180, 0.2350], + device='cuda:0'), in_proj_covar=tensor([0.0506, 0.0564, 0.0543, 0.0611, 0.0630, 0.0566, 0.0506, 0.0616], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 20:22:13,011 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0288, 1.6207, 1.4531, 1.5223, 1.3121, 1.3282, 1.2353, 1.2603], + device='cuda:0'), covar=tensor([0.1224, 0.0469, 0.1260, 0.0591, 0.0817, 0.1426, 0.0972, 0.0849], + device='cuda:0'), in_proj_covar=tensor([0.0348, 0.0229, 0.0320, 0.0297, 0.0293, 0.0326, 0.0337, 0.0308], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 20:22:16,307 INFO [train.py:901] (0/4) Epoch 17, batch 6100, loss[loss=0.2288, simple_loss=0.3043, pruned_loss=0.0766, over 8502.00 frames. ], tot_loss[loss=0.2139, simple_loss=0.2938, pruned_loss=0.06701, over 1610874.08 frames. ], batch size: 26, lr: 4.44e-03, grad_scale: 8.0 +2023-02-06 20:22:43,966 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.79 vs. limit=5.0 +2023-02-06 20:22:47,554 INFO [optim.py:369] (0/4) 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 20:22:49,612 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-06 20:22:52,249 INFO [train.py:901] (0/4) Epoch 17, batch 6150, loss[loss=0.205, simple_loss=0.278, pruned_loss=0.06597, over 7973.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.2954, pruned_loss=0.06798, over 1612943.66 frames. ], batch size: 21, lr: 4.44e-03, grad_scale: 8.0 +2023-02-06 20:23:02,061 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.83 vs. limit=5.0 +2023-02-06 20:23:26,579 INFO [train.py:901] (0/4) Epoch 17, batch 6200, loss[loss=0.1805, simple_loss=0.2517, pruned_loss=0.05461, over 7686.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.2949, pruned_loss=0.06734, over 1611711.68 frames. ], batch size: 18, lr: 4.44e-03, grad_scale: 8.0 +2023-02-06 20:23:29,342 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=135533.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:23:57,592 INFO [optim.py:369] (0/4) 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 20:24:01,718 INFO [train.py:901] (0/4) Epoch 17, batch 6250, loss[loss=0.2239, simple_loss=0.3192, pruned_loss=0.06434, over 8521.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.2951, pruned_loss=0.06702, over 1615135.39 frames. ], batch size: 28, lr: 4.44e-03, grad_scale: 8.0 +2023-02-06 20:24:13,143 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=135595.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:24:14,679 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.57 vs. limit=5.0 +2023-02-06 20:24:30,220 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=135620.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:24:36,941 INFO [train.py:901] (0/4) Epoch 17, batch 6300, loss[loss=0.2185, simple_loss=0.3054, pruned_loss=0.06582, over 8193.00 frames. ], tot_loss[loss=0.2144, simple_loss=0.2943, pruned_loss=0.06724, over 1609469.86 frames. ], batch size: 23, lr: 4.44e-03, grad_scale: 8.0 +2023-02-06 20:24:49,742 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=135648.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:24:49,779 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=135648.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:24:55,469 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 +2023-02-06 20:25:07,387 INFO [optim.py:369] (0/4) 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 20:25:11,440 INFO [train.py:901] (0/4) Epoch 17, batch 6350, loss[loss=0.2187, simple_loss=0.3004, pruned_loss=0.0685, over 8132.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.2947, pruned_loss=0.06719, over 1610981.14 frames. ], batch size: 22, lr: 4.44e-03, grad_scale: 8.0 +2023-02-06 20:25:15,104 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5259, 1.9773, 3.3052, 1.3193, 2.4643, 2.0244, 1.5992, 2.4630], + device='cuda:0'), covar=tensor([0.1883, 0.2414, 0.0688, 0.4329, 0.1635, 0.2833, 0.2128, 0.2083], + device='cuda:0'), in_proj_covar=tensor([0.0505, 0.0564, 0.0544, 0.0609, 0.0630, 0.0568, 0.0506, 0.0616], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 20:25:46,616 INFO [train.py:901] (0/4) Epoch 17, batch 6400, loss[loss=0.2023, simple_loss=0.2799, pruned_loss=0.06239, over 7445.00 frames. ], tot_loss[loss=0.2144, simple_loss=0.2942, pruned_loss=0.06725, over 1608456.32 frames. ], batch size: 17, lr: 4.43e-03, grad_scale: 8.0 +2023-02-06 20:25:52,203 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6228, 1.3275, 1.5070, 1.2677, 0.8166, 1.3387, 1.4397, 1.2989], + device='cuda:0'), covar=tensor([0.0536, 0.1323, 0.1741, 0.1501, 0.0608, 0.1576, 0.0731, 0.0674], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0152, 0.0190, 0.0157, 0.0100, 0.0162, 0.0114, 0.0139], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 20:26:16,675 INFO [optim.py:369] (0/4) 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 20:26:20,478 INFO [train.py:901] (0/4) Epoch 17, batch 6450, loss[loss=0.2025, simple_loss=0.293, pruned_loss=0.05597, over 8026.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.2942, pruned_loss=0.06765, over 1605777.41 frames. ], batch size: 22, lr: 4.43e-03, grad_scale: 8.0 +2023-02-06 20:26:56,417 INFO [train.py:901] (0/4) Epoch 17, batch 6500, loss[loss=0.2019, simple_loss=0.2868, pruned_loss=0.05855, over 8029.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2936, pruned_loss=0.06738, over 1602531.91 frames. ], batch size: 22, lr: 4.43e-03, grad_scale: 8.0 +2023-02-06 20:26:57,318 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8882, 2.2810, 3.8906, 1.7182, 2.9070, 2.3451, 1.9132, 2.8632], + device='cuda:0'), covar=tensor([0.1605, 0.2206, 0.0698, 0.3846, 0.1631, 0.2689, 0.1914, 0.2113], + device='cuda:0'), in_proj_covar=tensor([0.0508, 0.0568, 0.0546, 0.0614, 0.0634, 0.0573, 0.0508, 0.0619], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 20:27:26,781 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.9150, 5.9991, 5.2049, 2.5046, 5.3454, 5.6877, 5.4780, 5.3201], + device='cuda:0'), covar=tensor([0.0475, 0.0384, 0.0845, 0.4013, 0.0637, 0.0674, 0.0974, 0.0512], + device='cuda:0'), in_proj_covar=tensor([0.0503, 0.0416, 0.0420, 0.0519, 0.0409, 0.0413, 0.0402, 0.0362], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 20:27:27,349 INFO [optim.py:369] (0/4) 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 20:27:31,532 INFO [train.py:901] (0/4) Epoch 17, batch 6550, loss[loss=0.1975, simple_loss=0.2769, pruned_loss=0.05901, over 8460.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2922, pruned_loss=0.06611, over 1604763.30 frames. ], batch size: 29, lr: 4.43e-03, grad_scale: 8.0 +2023-02-06 20:27:48,288 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=135904.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:27:56,453 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-06 20:28:06,529 INFO [train.py:901] (0/4) Epoch 17, batch 6600, loss[loss=0.2399, simple_loss=0.3104, pruned_loss=0.08472, over 8233.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2938, pruned_loss=0.06733, over 1603826.63 frames. ], batch size: 22, lr: 4.43e-03, grad_scale: 8.0 +2023-02-06 20:28:06,736 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=135929.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:28:15,499 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.97 vs. limit=5.0 +2023-02-06 20:28:16,347 WARNING [train.py:1067] (0/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-06 20:28:19,938 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8010, 2.3532, 1.9753, 4.1991, 1.5767, 1.8001, 2.2220, 2.7320], + device='cuda:0'), covar=tensor([0.1566, 0.1296, 0.1821, 0.0228, 0.1457, 0.1793, 0.1246, 0.1075], + device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0196, 0.0247, 0.0210, 0.0207, 0.0245, 0.0253, 0.0209], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 20:28:24,643 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4120, 1.5161, 1.3767, 1.8559, 0.8991, 1.2529, 1.2878, 1.5094], + device='cuda:0'), covar=tensor([0.0842, 0.0773, 0.1047, 0.0499, 0.1081, 0.1414, 0.0824, 0.0737], + device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0196, 0.0247, 0.0210, 0.0207, 0.0245, 0.0253, 0.0209], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 20:28:36,539 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 17, batch 6650, loss[loss=0.2009, simple_loss=0.28, pruned_loss=0.06089, over 7975.00 frames. ], tot_loss[loss=0.2136, simple_loss=0.2932, pruned_loss=0.06702, over 1604057.24 frames. ], batch size: 21, lr: 4.43e-03, grad_scale: 8.0 +2023-02-06 20:28:49,875 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=135992.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:28:55,262 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-136000.pt +2023-02-06 20:28:56,423 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136000.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:29:04,534 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136012.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:29:16,453 INFO [train.py:901] (0/4) Epoch 17, batch 6700, loss[loss=0.2323, simple_loss=0.3133, pruned_loss=0.07561, over 8614.00 frames. ], tot_loss[loss=0.2161, simple_loss=0.2956, pruned_loss=0.06828, over 1610419.40 frames. ], batch size: 39, lr: 4.43e-03, grad_scale: 8.0 +2023-02-06 20:29:31,203 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.90 vs. limit=5.0 +2023-02-06 20:29:47,791 INFO [optim.py:369] (0/4) 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 20:29:51,767 INFO [train.py:901] (0/4) Epoch 17, batch 6750, loss[loss=0.1967, simple_loss=0.2804, pruned_loss=0.0565, over 7925.00 frames. ], tot_loss[loss=0.216, simple_loss=0.2949, pruned_loss=0.06849, over 1606978.33 frames. ], batch size: 20, lr: 4.43e-03, grad_scale: 8.0 +2023-02-06 20:30:11,742 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136107.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:30:26,349 INFO [train.py:901] (0/4) Epoch 17, batch 6800, loss[loss=0.1934, simple_loss=0.2811, pruned_loss=0.0528, over 8028.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2944, pruned_loss=0.06789, over 1610127.74 frames. ], batch size: 22, lr: 4.43e-03, grad_scale: 8.0 +2023-02-06 20:30:35,096 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-06 20:30:39,133 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.6951, 4.7671, 4.2501, 2.1110, 4.0957, 4.2732, 4.3385, 4.0182], + device='cuda:0'), covar=tensor([0.0728, 0.0475, 0.0977, 0.4812, 0.0921, 0.1068, 0.1112, 0.0796], + device='cuda:0'), in_proj_covar=tensor([0.0506, 0.0417, 0.0421, 0.0523, 0.0412, 0.0416, 0.0404, 0.0364], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 20:30:48,379 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 20:31:01,822 INFO [train.py:901] (0/4) Epoch 17, batch 6850, loss[loss=0.2061, simple_loss=0.2896, pruned_loss=0.06128, over 8359.00 frames. ], tot_loss[loss=0.2153, simple_loss=0.2948, pruned_loss=0.06787, over 1609354.59 frames. ], batch size: 24, lr: 4.43e-03, grad_scale: 8.0 +2023-02-06 20:31:23,505 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-06 20:31:37,306 INFO [train.py:901] (0/4) Epoch 17, batch 6900, loss[loss=0.2019, simple_loss=0.2909, pruned_loss=0.05646, over 8361.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2943, pruned_loss=0.06775, over 1608364.26 frames. ], batch size: 24, lr: 4.43e-03, grad_scale: 8.0 +2023-02-06 20:31:59,827 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.7667, 1.7406, 1.8237, 1.7451, 0.9642, 1.6465, 2.0533, 2.1013], + device='cuda:0'), covar=tensor([0.0425, 0.1245, 0.1730, 0.1331, 0.0621, 0.1496, 0.0675, 0.0552], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0154, 0.0192, 0.0159, 0.0101, 0.0164, 0.0115, 0.0140], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 20:32:03,281 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3164, 1.3037, 1.4710, 1.2777, 0.7665, 1.2642, 1.2389, 1.1828], + device='cuda:0'), covar=tensor([0.0558, 0.1392, 0.1800, 0.1476, 0.0586, 0.1676, 0.0760, 0.0659], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0154, 0.0192, 0.0159, 0.0101, 0.0164, 0.0115, 0.0140], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 20:32:08,491 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 17, batch 6950, loss[loss=0.1753, simple_loss=0.2556, pruned_loss=0.04748, over 7426.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.2944, pruned_loss=0.06738, over 1607237.50 frames. ], batch size: 17, lr: 4.43e-03, grad_scale: 8.0 +2023-02-06 20:32:32,967 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-06 20:32:33,817 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.7231, 2.7486, 2.4312, 4.0286, 1.6040, 2.1433, 2.5385, 2.9015], + device='cuda:0'), covar=tensor([0.0623, 0.0877, 0.0893, 0.0226, 0.1266, 0.1198, 0.1020, 0.0810], + device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0197, 0.0248, 0.0212, 0.0209, 0.0247, 0.0254, 0.0209], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 20:32:48,141 INFO [train.py:901] (0/4) Epoch 17, batch 7000, loss[loss=0.2107, simple_loss=0.292, pruned_loss=0.06475, over 8078.00 frames. ], tot_loss[loss=0.2138, simple_loss=0.2934, pruned_loss=0.06707, over 1602201.29 frames. ], batch size: 21, lr: 4.43e-03, grad_scale: 8.0 +2023-02-06 20:32:50,919 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5615, 2.6401, 1.7598, 2.1329, 2.1572, 1.5810, 1.9849, 2.1492], + device='cuda:0'), covar=tensor([0.1346, 0.0325, 0.1137, 0.0640, 0.0661, 0.1348, 0.0998, 0.0977], + device='cuda:0'), in_proj_covar=tensor([0.0349, 0.0229, 0.0321, 0.0297, 0.0294, 0.0327, 0.0339, 0.0310], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 20:32:58,258 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=136344.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:33:04,320 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136353.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:33:06,280 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=136356.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:33:06,417 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9673, 3.6446, 2.2055, 2.7193, 2.9168, 1.9951, 2.7508, 2.9396], + device='cuda:0'), covar=tensor([0.1655, 0.0342, 0.1173, 0.0865, 0.0718, 0.1450, 0.1068, 0.1159], + device='cuda:0'), in_proj_covar=tensor([0.0349, 0.0230, 0.0321, 0.0297, 0.0294, 0.0327, 0.0339, 0.0310], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 20:33:11,077 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 20:33:22,385 INFO [train.py:901] (0/4) Epoch 17, batch 7050, loss[loss=0.2266, simple_loss=0.3107, pruned_loss=0.07131, over 8491.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.2947, pruned_loss=0.06742, over 1611136.24 frames. ], batch size: 29, lr: 4.42e-03, grad_scale: 8.0 +2023-02-06 20:33:29,397 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=136388.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:33:57,957 INFO [train.py:901] (0/4) Epoch 17, batch 7100, loss[loss=0.2229, simple_loss=0.3074, pruned_loss=0.06916, over 8028.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.2954, pruned_loss=0.06752, over 1616621.89 frames. ], batch size: 22, lr: 4.42e-03, grad_scale: 8.0 +2023-02-06 20:34:18,665 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136459.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:34:26,514 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136471.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:34:27,625 INFO [optim.py:369] (0/4) 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 20:34:31,738 INFO [train.py:901] (0/4) Epoch 17, batch 7150, loss[loss=0.2242, simple_loss=0.3121, pruned_loss=0.06819, over 8194.00 frames. ], tot_loss[loss=0.2161, simple_loss=0.2962, pruned_loss=0.06802, over 1614496.16 frames. ], batch size: 23, lr: 4.42e-03, grad_scale: 8.0 +2023-02-06 20:34:50,070 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=136504.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:35:07,634 INFO [train.py:901] (0/4) Epoch 17, batch 7200, loss[loss=0.2191, simple_loss=0.3058, pruned_loss=0.06622, over 8484.00 frames. ], tot_loss[loss=0.2154, simple_loss=0.2958, pruned_loss=0.06753, over 1617024.22 frames. ], batch size: 29, lr: 4.42e-03, grad_scale: 8.0 +2023-02-06 20:35:37,951 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 17, batch 7250, loss[loss=0.1899, simple_loss=0.2789, pruned_loss=0.05051, over 8574.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.2965, pruned_loss=0.06797, over 1622541.04 frames. ], batch size: 31, lr: 4.42e-03, grad_scale: 8.0 +2023-02-06 20:36:11,358 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136619.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:36:17,802 INFO [train.py:901] (0/4) Epoch 17, batch 7300, loss[loss=0.2182, simple_loss=0.3052, pruned_loss=0.06561, over 8103.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2963, pruned_loss=0.06764, over 1616800.07 frames. ], batch size: 23, lr: 4.42e-03, grad_scale: 8.0 +2023-02-06 20:36:40,597 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136661.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:36:42,653 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0160, 1.5313, 3.5215, 1.5558, 2.3176, 3.9386, 3.9451, 3.3821], + device='cuda:0'), covar=tensor([0.1110, 0.1796, 0.0348, 0.2099, 0.1152, 0.0193, 0.0512, 0.0518], + device='cuda:0'), in_proj_covar=tensor([0.0285, 0.0314, 0.0278, 0.0307, 0.0296, 0.0255, 0.0392, 0.0299], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-06 20:36:48,514 INFO [optim.py:369] (0/4) 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 20:36:52,586 INFO [train.py:901] (0/4) Epoch 17, batch 7350, loss[loss=0.2509, simple_loss=0.3191, pruned_loss=0.09138, over 8252.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.2955, pruned_loss=0.06702, over 1615824.14 frames. ], batch size: 22, lr: 4.42e-03, grad_scale: 8.0 +2023-02-06 20:37:05,619 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=136697.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:37:15,662 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 +2023-02-06 20:37:16,545 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-06 20:37:17,974 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136715.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:37:27,008 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136727.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:37:28,131 INFO [train.py:901] (0/4) Epoch 17, batch 7400, loss[loss=0.2534, simple_loss=0.3306, pruned_loss=0.08808, over 8284.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.2949, pruned_loss=0.06716, over 1607374.04 frames. ], batch size: 23, lr: 4.42e-03, grad_scale: 8.0 +2023-02-06 20:37:35,006 WARNING [train.py:1067] (0/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-06 20:37:36,558 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=136740.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:37:45,373 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=136752.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:37:59,284 INFO [optim.py:369] (0/4) 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 20:38:03,291 INFO [train.py:901] (0/4) Epoch 17, batch 7450, loss[loss=0.1753, simple_loss=0.2537, pruned_loss=0.04844, over 7419.00 frames. ], tot_loss[loss=0.215, simple_loss=0.2953, pruned_loss=0.06735, over 1610913.12 frames. ], batch size: 17, lr: 4.42e-03, grad_scale: 8.0 +2023-02-06 20:38:16,561 WARNING [train.py:1067] (0/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-06 20:38:26,046 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136812.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:38:34,160 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136824.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:38:37,329 INFO [train.py:901] (0/4) Epoch 17, batch 7500, loss[loss=0.2361, simple_loss=0.3188, pruned_loss=0.07667, over 8195.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.2957, pruned_loss=0.06765, over 1609160.18 frames. ], batch size: 23, lr: 4.42e-03, grad_scale: 8.0 +2023-02-06 20:39:09,456 INFO [optim.py:369] (0/4) 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 20:39:11,055 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 17, batch 7550, loss[loss=0.1711, simple_loss=0.2537, pruned_loss=0.04427, over 7692.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2931, pruned_loss=0.06692, over 1599516.17 frames. ], batch size: 18, lr: 4.42e-03, grad_scale: 8.0 +2023-02-06 20:39:28,588 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=136900.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:39:49,026 INFO [train.py:901] (0/4) Epoch 17, batch 7600, loss[loss=0.2479, simple_loss=0.334, pruned_loss=0.08087, over 8587.00 frames. ], tot_loss[loss=0.2126, simple_loss=0.2925, pruned_loss=0.06633, over 1599037.77 frames. ], batch size: 31, lr: 4.42e-03, grad_scale: 8.0 +2023-02-06 20:40:14,539 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0500, 2.3573, 3.3596, 1.8705, 2.7329, 2.3282, 2.0831, 2.6386], + device='cuda:0'), covar=tensor([0.1402, 0.1942, 0.0556, 0.3345, 0.1364, 0.2317, 0.1682, 0.1828], + device='cuda:0'), in_proj_covar=tensor([0.0509, 0.0571, 0.0545, 0.0614, 0.0635, 0.0576, 0.0509, 0.0619], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 20:40:21,096 INFO [optim.py:369] (0/4) 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 20:40:22,731 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2925, 2.1120, 2.8431, 2.3067, 2.7366, 2.2978, 2.0165, 1.6073], + device='cuda:0'), covar=tensor([0.4675, 0.4386, 0.1543, 0.3079, 0.2215, 0.2620, 0.1690, 0.4786], + device='cuda:0'), in_proj_covar=tensor([0.0920, 0.0937, 0.0775, 0.0902, 0.0969, 0.0853, 0.0723, 0.0798], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 20:40:25,199 INFO [train.py:901] (0/4) Epoch 17, batch 7650, loss[loss=0.2706, simple_loss=0.3304, pruned_loss=0.1054, over 7914.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2924, pruned_loss=0.06646, over 1599219.71 frames. ], batch size: 20, lr: 4.41e-03, grad_scale: 16.0 +2023-02-06 20:40:38,666 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4813, 1.8286, 2.7257, 1.3217, 2.0220, 1.7825, 1.5785, 2.0136], + device='cuda:0'), covar=tensor([0.1814, 0.2326, 0.0789, 0.4276, 0.1686, 0.3013, 0.2088, 0.2080], + device='cuda:0'), in_proj_covar=tensor([0.0510, 0.0572, 0.0546, 0.0615, 0.0636, 0.0577, 0.0509, 0.0620], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 20:40:43,288 INFO [zipformer.py:1185] (0/4) 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] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-02-06 20:41:00,039 INFO [train.py:901] (0/4) Epoch 17, batch 7700, loss[loss=0.2552, simple_loss=0.3135, pruned_loss=0.09848, over 6696.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.2942, pruned_loss=0.06767, over 1600368.31 frames. ], batch size: 71, lr: 4.41e-03, grad_scale: 16.0 +2023-02-06 20:41:26,489 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-06 20:41:26,683 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=137068.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:41:30,554 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 17, batch 7750, loss[loss=0.1935, simple_loss=0.2763, pruned_loss=0.05539, over 7817.00 frames. ], tot_loss[loss=0.2166, simple_loss=0.2957, pruned_loss=0.0688, over 1602813.05 frames. ], batch size: 20, lr: 4.41e-03, grad_scale: 16.0 +2023-02-06 20:41:35,626 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.1191, 2.5354, 3.7274, 2.1807, 1.9172, 3.7781, 0.7562, 2.2263], + device='cuda:0'), covar=tensor([0.1350, 0.1257, 0.0241, 0.1630, 0.2864, 0.0254, 0.2481, 0.1585], + device='cuda:0'), in_proj_covar=tensor([0.0176, 0.0183, 0.0116, 0.0216, 0.0263, 0.0123, 0.0166, 0.0182], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 20:41:44,964 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=137093.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:42:03,591 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=137120.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:42:09,483 INFO [train.py:901] (0/4) Epoch 17, batch 7800, loss[loss=0.1967, simple_loss=0.2692, pruned_loss=0.06209, over 7186.00 frames. ], tot_loss[loss=0.2153, simple_loss=0.2945, pruned_loss=0.06803, over 1601936.09 frames. ], batch size: 16, lr: 4.41e-03, grad_scale: 16.0 +2023-02-06 20:42:36,394 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=137168.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:42:39,569 INFO [optim.py:369] (0/4) 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 20:42:41,430 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 +2023-02-06 20:42:43,610 INFO [train.py:901] (0/4) Epoch 17, batch 7850, loss[loss=0.2135, simple_loss=0.2968, pruned_loss=0.0651, over 8567.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2958, pruned_loss=0.0685, over 1605093.62 frames. ], batch size: 49, lr: 4.41e-03, grad_scale: 16.0 +2023-02-06 20:43:16,598 INFO [train.py:901] (0/4) Epoch 17, batch 7900, loss[loss=0.206, simple_loss=0.292, pruned_loss=0.05999, over 8616.00 frames. ], tot_loss[loss=0.2176, simple_loss=0.2968, pruned_loss=0.06921, over 1606746.26 frames. ], batch size: 31, lr: 4.41e-03, grad_scale: 16.0 +2023-02-06 20:43:45,782 INFO [optim.py:369] (0/4) 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 20:43:49,859 INFO [train.py:901] (0/4) Epoch 17, batch 7950, loss[loss=0.2309, simple_loss=0.3159, pruned_loss=0.0729, over 8587.00 frames. ], tot_loss[loss=0.217, simple_loss=0.2963, pruned_loss=0.0688, over 1611235.78 frames. ], batch size: 31, lr: 4.41e-03, grad_scale: 16.0 +2023-02-06 20:43:52,821 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=137283.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:44:23,210 INFO [train.py:901] (0/4) Epoch 17, batch 8000, loss[loss=0.1696, simple_loss=0.2467, pruned_loss=0.04625, over 7222.00 frames. ], tot_loss[loss=0.215, simple_loss=0.2942, pruned_loss=0.06794, over 1602915.68 frames. ], batch size: 16, lr: 4.41e-03, grad_scale: 16.0 +2023-02-06 20:44:45,095 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0348, 1.7013, 3.0557, 1.3651, 2.2177, 3.3658, 3.4638, 2.7876], + device='cuda:0'), covar=tensor([0.1028, 0.1556, 0.0373, 0.2137, 0.1032, 0.0260, 0.0499, 0.0667], + device='cuda:0'), in_proj_covar=tensor([0.0284, 0.0313, 0.0277, 0.0306, 0.0297, 0.0254, 0.0392, 0.0298], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-06 20:44:52,972 INFO [optim.py:369] (0/4) 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] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=137376.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:44:57,243 INFO [train.py:901] (0/4) Epoch 17, batch 8050, loss[loss=0.1725, simple_loss=0.2423, pruned_loss=0.05138, over 7420.00 frames. ], tot_loss[loss=0.2153, simple_loss=0.2944, pruned_loss=0.06812, over 1595580.10 frames. ], batch size: 17, lr: 4.41e-03, grad_scale: 16.0 +2023-02-06 20:45:12,511 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=137401.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:45:20,315 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/epoch-17.pt +2023-02-06 20:45:31,418 WARNING [train.py:1067] (0/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-06 20:45:34,953 INFO [train.py:901] (0/4) Epoch 18, batch 0, loss[loss=0.2083, simple_loss=0.2966, pruned_loss=0.05997, over 8297.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2966, pruned_loss=0.05997, over 8297.00 frames. ], batch size: 23, lr: 4.28e-03, grad_scale: 16.0 +2023-02-06 20:45:34,954 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 20:45:46,126 INFO [train.py:935] (0/4) Epoch 18, validation: loss=0.1783, simple_loss=0.2784, pruned_loss=0.03907, over 944034.00 frames. +2023-02-06 20:45:46,127 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-06 20:46:00,879 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-06 20:46:20,803 INFO [train.py:901] (0/4) Epoch 18, batch 50, loss[loss=0.1906, simple_loss=0.2711, pruned_loss=0.05502, over 7929.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.2963, pruned_loss=0.06928, over 368167.69 frames. ], batch size: 20, lr: 4.28e-03, grad_scale: 16.0 +2023-02-06 20:46:25,301 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 +2023-02-06 20:46:29,001 INFO [optim.py:369] (0/4) 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 20:46:35,868 WARNING [train.py:1067] (0/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-06 20:46:47,477 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6009, 1.4433, 1.6593, 1.3192, 0.9016, 1.4230, 1.4555, 1.2451], + device='cuda:0'), covar=tensor([0.0581, 0.1246, 0.1629, 0.1445, 0.0604, 0.1487, 0.0754, 0.0697], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0152, 0.0190, 0.0157, 0.0100, 0.0161, 0.0113, 0.0138], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0008, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 20:46:56,046 INFO [train.py:901] (0/4) Epoch 18, batch 100, loss[loss=0.1845, simple_loss=0.2648, pruned_loss=0.05212, over 7277.00 frames. ], tot_loss[loss=0.2189, simple_loss=0.2974, pruned_loss=0.07025, over 641071.54 frames. ], batch size: 16, lr: 4.28e-03, grad_scale: 16.0 +2023-02-06 20:46:58,829 WARNING [train.py:1067] (0/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-06 20:47:16,414 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 18, batch 150, loss[loss=0.2116, simple_loss=0.2943, pruned_loss=0.06448, over 8385.00 frames. ], tot_loss[loss=0.2184, simple_loss=0.2971, pruned_loss=0.06979, over 861937.76 frames. ], batch size: 49, lr: 4.28e-03, grad_scale: 16.0 +2023-02-06 20:47:33,465 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=137564.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:47:39,685 INFO [optim.py:369] (0/4) 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 20:48:07,683 INFO [train.py:901] (0/4) Epoch 18, batch 200, loss[loss=0.2044, simple_loss=0.2846, pruned_loss=0.06214, over 8229.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.2952, pruned_loss=0.06785, over 1033345.98 frames. ], batch size: 22, lr: 4.28e-03, grad_scale: 16.0 +2023-02-06 20:48:44,090 INFO [train.py:901] (0/4) Epoch 18, batch 250, loss[loss=0.2481, simple_loss=0.3166, pruned_loss=0.08976, over 8503.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.2943, pruned_loss=0.06746, over 1160760.76 frames. ], batch size: 29, lr: 4.28e-03, grad_scale: 16.0 +2023-02-06 20:48:52,403 INFO [optim.py:369] (0/4) 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 20:48:55,938 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-06 20:49:03,706 WARNING [train.py:1067] (0/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-06 20:49:19,874 INFO [train.py:901] (0/4) Epoch 18, batch 300, loss[loss=0.199, simple_loss=0.2776, pruned_loss=0.06018, over 7814.00 frames. ], tot_loss[loss=0.216, simple_loss=0.2952, pruned_loss=0.06838, over 1256585.67 frames. ], batch size: 20, lr: 4.28e-03, grad_scale: 8.0 +2023-02-06 20:49:31,576 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.85 vs. limit=5.0 +2023-02-06 20:49:55,780 INFO [train.py:901] (0/4) Epoch 18, batch 350, loss[loss=0.2331, simple_loss=0.3124, pruned_loss=0.07691, over 8468.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2945, pruned_loss=0.06791, over 1333732.47 frames. ], batch size: 25, lr: 4.28e-03, grad_scale: 8.0 +2023-02-06 20:50:03,812 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7588, 1.8952, 1.7240, 2.3141, 0.9502, 1.4410, 1.6068, 1.9196], + device='cuda:0'), covar=tensor([0.0778, 0.0687, 0.0993, 0.0443, 0.1135, 0.1352, 0.0881, 0.0665], + device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0198, 0.0251, 0.0211, 0.0208, 0.0249, 0.0255, 0.0209], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 20:50:05,700 INFO [optim.py:369] (0/4) 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 20:50:32,304 INFO [train.py:901] (0/4) Epoch 18, batch 400, loss[loss=0.1919, simple_loss=0.2649, pruned_loss=0.05946, over 7689.00 frames. ], tot_loss[loss=0.2154, simple_loss=0.2952, pruned_loss=0.06784, over 1397234.62 frames. ], batch size: 18, lr: 4.28e-03, grad_scale: 8.0 +2023-02-06 20:51:08,176 INFO [train.py:901] (0/4) Epoch 18, batch 450, loss[loss=0.2159, simple_loss=0.2867, pruned_loss=0.07253, over 7661.00 frames. ], tot_loss[loss=0.217, simple_loss=0.2967, pruned_loss=0.06867, over 1448360.98 frames. ], batch size: 19, lr: 4.27e-03, grad_scale: 8.0 +2023-02-06 20:51:16,912 INFO [optim.py:369] (0/4) 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 20:51:37,225 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0891, 1.5367, 3.5360, 1.3373, 2.4939, 3.9236, 3.9739, 3.3155], + device='cuda:0'), covar=tensor([0.1138, 0.1704, 0.0323, 0.2155, 0.0953, 0.0214, 0.0546, 0.0664], + device='cuda:0'), in_proj_covar=tensor([0.0284, 0.0314, 0.0277, 0.0307, 0.0296, 0.0254, 0.0396, 0.0299], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-06 20:51:43,032 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=137910.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:51:43,588 INFO [train.py:901] (0/4) Epoch 18, batch 500, loss[loss=0.1753, simple_loss=0.2572, pruned_loss=0.04667, over 7645.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2965, pruned_loss=0.06815, over 1483310.26 frames. ], batch size: 19, lr: 4.27e-03, grad_scale: 8.0 +2023-02-06 20:51:45,164 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=137913.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:51:46,137 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 +2023-02-06 20:52:04,555 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.3816, 4.3899, 3.9576, 1.9279, 3.8563, 3.9788, 3.8910, 3.7512], + device='cuda:0'), covar=tensor([0.0820, 0.0583, 0.1149, 0.5315, 0.0981, 0.1075, 0.1375, 0.0886], + device='cuda:0'), in_proj_covar=tensor([0.0509, 0.0418, 0.0419, 0.0524, 0.0414, 0.0421, 0.0410, 0.0366], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 20:52:20,558 INFO [train.py:901] (0/4) Epoch 18, batch 550, loss[loss=0.2522, simple_loss=0.3064, pruned_loss=0.09901, over 7695.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.2971, pruned_loss=0.06825, over 1516881.09 frames. ], batch size: 18, lr: 4.27e-03, grad_scale: 8.0 +2023-02-06 20:52:29,488 INFO [optim.py:369] (0/4) 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 20:52:48,226 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-138000.pt +2023-02-06 20:52:56,990 INFO [train.py:901] (0/4) Epoch 18, batch 600, loss[loss=0.2424, simple_loss=0.3128, pruned_loss=0.08598, over 8039.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2966, pruned_loss=0.0681, over 1540780.49 frames. ], batch size: 22, lr: 4.27e-03, grad_scale: 8.0 +2023-02-06 20:53:09,283 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1290, 2.3163, 1.9904, 2.9362, 1.4235, 1.6189, 2.0013, 2.3511], + device='cuda:0'), covar=tensor([0.0722, 0.0739, 0.0978, 0.0365, 0.1131, 0.1407, 0.0976, 0.0718], + device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0199, 0.0251, 0.0212, 0.0208, 0.0249, 0.0254, 0.0210], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 20:53:11,919 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-06 20:53:32,788 INFO [train.py:901] (0/4) Epoch 18, batch 650, loss[loss=0.185, simple_loss=0.2728, pruned_loss=0.04857, over 8187.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.2959, pruned_loss=0.06721, over 1562139.86 frames. ], batch size: 23, lr: 4.27e-03, grad_scale: 8.0 +2023-02-06 20:53:43,406 INFO [optim.py:369] (0/4) 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 20:54:00,542 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-06 20:54:09,454 INFO [train.py:901] (0/4) Epoch 18, batch 700, loss[loss=0.2982, simple_loss=0.3556, pruned_loss=0.1204, over 8734.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.2954, pruned_loss=0.06706, over 1575970.70 frames. ], batch size: 30, lr: 4.27e-03, grad_scale: 8.0 +2023-02-06 20:54:29,540 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 +2023-02-06 20:54:30,784 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5624, 1.4561, 1.9103, 1.4025, 1.1548, 2.0018, 0.4240, 1.3449], + device='cuda:0'), covar=tensor([0.2046, 0.1472, 0.0418, 0.1135, 0.3027, 0.0426, 0.2348, 0.1489], + device='cuda:0'), in_proj_covar=tensor([0.0178, 0.0184, 0.0117, 0.0217, 0.0262, 0.0123, 0.0165, 0.0182], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 20:54:44,125 INFO [train.py:901] (0/4) Epoch 18, batch 750, loss[loss=0.1731, simple_loss=0.2553, pruned_loss=0.04546, over 7648.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.2959, pruned_loss=0.06722, over 1586254.21 frames. ], batch size: 19, lr: 4.27e-03, grad_scale: 8.0 +2023-02-06 20:54:53,227 INFO [optim.py:369] (0/4) 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 20:54:58,215 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-06 20:55:08,073 WARNING [train.py:1067] (0/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-06 20:55:19,810 INFO [train.py:901] (0/4) Epoch 18, batch 800, loss[loss=0.2765, simple_loss=0.3426, pruned_loss=0.1052, over 7043.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2953, pruned_loss=0.06728, over 1590608.88 frames. ], batch size: 71, lr: 4.27e-03, grad_scale: 8.0 +2023-02-06 20:55:49,573 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=138254.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:55:51,615 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=138257.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:55:54,239 INFO [train.py:901] (0/4) Epoch 18, batch 850, loss[loss=0.2033, simple_loss=0.2824, pruned_loss=0.06208, over 8254.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2937, pruned_loss=0.06662, over 1597850.90 frames. ], batch size: 22, lr: 4.27e-03, grad_scale: 8.0 +2023-02-06 20:56:03,039 INFO [optim.py:369] (0/4) 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 20:56:13,562 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138288.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:56:30,839 INFO [train.py:901] (0/4) Epoch 18, batch 900, loss[loss=0.1933, simple_loss=0.2826, pruned_loss=0.05198, over 7544.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2919, pruned_loss=0.06523, over 1606910.18 frames. ], batch size: 18, lr: 4.27e-03, grad_scale: 8.0 +2023-02-06 20:56:35,138 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0433, 1.4648, 1.7584, 1.4382, 1.0666, 1.4693, 1.7655, 1.4902], + device='cuda:0'), covar=tensor([0.0486, 0.1302, 0.1626, 0.1389, 0.0575, 0.1504, 0.0698, 0.0665], + device='cuda:0'), in_proj_covar=tensor([0.0098, 0.0152, 0.0190, 0.0158, 0.0100, 0.0161, 0.0114, 0.0139], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 20:56:50,624 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138340.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:57:05,367 INFO [train.py:901] (0/4) Epoch 18, batch 950, loss[loss=0.1708, simple_loss=0.2484, pruned_loss=0.04663, over 7799.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2922, pruned_loss=0.06579, over 1608686.45 frames. ], batch size: 19, lr: 4.27e-03, grad_scale: 8.0 +2023-02-06 20:57:10,952 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138369.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:57:13,023 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138372.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 20:57:14,181 INFO [optim.py:369] (0/4) 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 20:57:14,401 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6542, 1.8591, 1.6989, 2.3290, 1.0669, 1.4379, 1.7590, 1.9944], + device='cuda:0'), covar=tensor([0.0833, 0.0728, 0.0931, 0.0435, 0.1034, 0.1310, 0.0783, 0.0623], + device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0200, 0.0251, 0.0212, 0.0207, 0.0249, 0.0253, 0.0208], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 20:57:29,245 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-06 20:57:40,362 INFO [train.py:901] (0/4) Epoch 18, batch 1000, loss[loss=0.2197, simple_loss=0.3004, pruned_loss=0.06954, over 8130.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2919, pruned_loss=0.06579, over 1610532.01 frames. ], batch size: 22, lr: 4.27e-03, grad_scale: 8.0 +2023-02-06 20:57:56,185 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8764, 2.2813, 4.2349, 1.4917, 2.8723, 2.3451, 1.8150, 2.8636], + device='cuda:0'), covar=tensor([0.1783, 0.2574, 0.0778, 0.4329, 0.1903, 0.2978, 0.2146, 0.2445], + device='cuda:0'), in_proj_covar=tensor([0.0511, 0.0574, 0.0544, 0.0618, 0.0636, 0.0577, 0.0510, 0.0624], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 20:58:05,494 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-06 20:58:16,701 INFO [train.py:901] (0/4) Epoch 18, batch 1050, loss[loss=0.2143, simple_loss=0.3008, pruned_loss=0.0639, over 8632.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.292, pruned_loss=0.06566, over 1610606.03 frames. ], batch size: 39, lr: 4.27e-03, grad_scale: 8.0 +2023-02-06 20:58:18,830 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-06 20:58:25,548 INFO [optim.py:369] (0/4) 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 20:58:51,053 INFO [train.py:901] (0/4) Epoch 18, batch 1100, loss[loss=0.2053, simple_loss=0.2846, pruned_loss=0.06298, over 7781.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2916, pruned_loss=0.06538, over 1616941.07 frames. ], batch size: 19, lr: 4.26e-03, grad_scale: 8.0 +2023-02-06 20:59:26,910 INFO [train.py:901] (0/4) Epoch 18, batch 1150, loss[loss=0.2138, simple_loss=0.3002, pruned_loss=0.06363, over 8507.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2914, pruned_loss=0.06557, over 1616965.93 frames. ], batch size: 28, lr: 4.26e-03, grad_scale: 8.0 +2023-02-06 20:59:29,613 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-06 20:59:35,876 INFO [optim.py:369] (0/4) 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 21:00:02,037 INFO [train.py:901] (0/4) Epoch 18, batch 1200, loss[loss=0.1871, simple_loss=0.2815, pruned_loss=0.04641, over 8322.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2922, pruned_loss=0.06577, over 1616513.89 frames. ], batch size: 26, lr: 4.26e-03, grad_scale: 8.0 +2023-02-06 21:00:11,833 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138625.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:00:13,958 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138628.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:00:16,607 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=138632.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:00:29,702 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138650.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:00:31,771 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 18, batch 1250, loss[loss=0.2892, simple_loss=0.3631, pruned_loss=0.1076, over 8460.00 frames. ], tot_loss[loss=0.2116, simple_loss=0.2925, pruned_loss=0.0654, over 1618238.13 frames. ], batch size: 27, lr: 4.26e-03, grad_scale: 8.0 +2023-02-06 21:00:47,269 INFO [optim.py:369] (0/4) 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 21:00:50,839 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.2263, 1.1655, 3.3267, 1.0313, 2.9540, 2.7717, 3.0499, 2.9508], + device='cuda:0'), covar=tensor([0.0799, 0.4354, 0.0944, 0.4464, 0.1433, 0.1174, 0.0790, 0.0939], + device='cuda:0'), in_proj_covar=tensor([0.0583, 0.0620, 0.0663, 0.0593, 0.0675, 0.0579, 0.0571, 0.0644], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 21:00:54,239 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=138684.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:01:04,822 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138699.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:01:08,240 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138704.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:01:12,848 INFO [train.py:901] (0/4) Epoch 18, batch 1300, loss[loss=0.2094, simple_loss=0.2975, pruned_loss=0.06067, over 8030.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2922, pruned_loss=0.06576, over 1617362.32 frames. ], batch size: 22, lr: 4.26e-03, grad_scale: 8.0 +2023-02-06 21:01:15,188 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0061, 3.5555, 2.3434, 2.9156, 2.7283, 2.1395, 2.7012, 3.0447], + device='cuda:0'), covar=tensor([0.1671, 0.0391, 0.1044, 0.0732, 0.0744, 0.1350, 0.1097, 0.1081], + device='cuda:0'), in_proj_covar=tensor([0.0351, 0.0236, 0.0327, 0.0302, 0.0299, 0.0331, 0.0345, 0.0316], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 21:01:15,823 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138715.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:01:21,888 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6244, 1.5713, 2.0947, 1.4931, 1.1480, 2.0978, 0.2592, 1.2995], + device='cuda:0'), covar=tensor([0.1899, 0.1468, 0.0386, 0.1207, 0.3302, 0.0395, 0.2372, 0.1516], + device='cuda:0'), in_proj_covar=tensor([0.0179, 0.0186, 0.0117, 0.0217, 0.0263, 0.0125, 0.0164, 0.0183], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 21:01:37,536 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138747.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:01:47,842 INFO [train.py:901] (0/4) Epoch 18, batch 1350, loss[loss=0.2328, simple_loss=0.2899, pruned_loss=0.08779, over 7224.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2925, pruned_loss=0.06578, over 1618891.65 frames. ], batch size: 16, lr: 4.26e-03, grad_scale: 8.0 +2023-02-06 21:01:49,487 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8484, 2.1866, 3.4081, 1.4124, 2.5689, 2.1407, 1.8337, 2.5471], + device='cuda:0'), covar=tensor([0.1700, 0.2387, 0.0815, 0.4252, 0.1728, 0.2967, 0.2088, 0.2214], + device='cuda:0'), in_proj_covar=tensor([0.0514, 0.0578, 0.0550, 0.0624, 0.0643, 0.0583, 0.0513, 0.0630], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 21:01:56,591 INFO [optim.py:369] (0/4) 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 21:02:15,433 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138799.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:02:23,333 INFO [train.py:901] (0/4) Epoch 18, batch 1400, loss[loss=0.2075, simple_loss=0.2877, pruned_loss=0.06368, over 8313.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.2933, pruned_loss=0.06654, over 1618369.96 frames. ], batch size: 25, lr: 4.26e-03, grad_scale: 8.0 +2023-02-06 21:02:49,883 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 +2023-02-06 21:02:54,708 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 +2023-02-06 21:02:57,617 INFO [train.py:901] (0/4) Epoch 18, batch 1450, loss[loss=0.2178, simple_loss=0.2969, pruned_loss=0.06937, over 8229.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.2926, pruned_loss=0.06611, over 1616808.80 frames. ], batch size: 22, lr: 4.26e-03, grad_scale: 8.0 +2023-02-06 21:03:06,393 INFO [optim.py:369] (0/4) 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 21:03:07,083 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-06 21:03:33,395 INFO [train.py:901] (0/4) Epoch 18, batch 1500, loss[loss=0.251, simple_loss=0.3267, pruned_loss=0.08764, over 8696.00 frames. ], tot_loss[loss=0.214, simple_loss=0.2941, pruned_loss=0.06699, over 1617785.02 frames. ], batch size: 39, lr: 4.26e-03, grad_scale: 8.0 +2023-02-06 21:03:36,675 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.09 vs. limit=5.0 +2023-02-06 21:03:44,747 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138927.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:04:04,115 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-06 21:04:08,387 INFO [train.py:901] (0/4) Epoch 18, batch 1550, loss[loss=0.2015, simple_loss=0.2833, pruned_loss=0.05989, over 7967.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2941, pruned_loss=0.06715, over 1617085.87 frames. ], batch size: 21, lr: 4.26e-03, grad_scale: 8.0 +2023-02-06 21:04:17,331 INFO [optim.py:369] (0/4) 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 21:04:38,171 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 18, batch 1600, loss[loss=0.1798, simple_loss=0.2585, pruned_loss=0.05052, over 7421.00 frames. ], tot_loss[loss=0.2138, simple_loss=0.294, pruned_loss=0.06684, over 1620138.62 frames. ], batch size: 17, lr: 4.26e-03, grad_scale: 8.0 +2023-02-06 21:04:56,938 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139028.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:04:59,592 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.5837, 4.5936, 4.0778, 1.8982, 4.0217, 4.1122, 4.1261, 3.9167], + device='cuda:0'), covar=tensor([0.0706, 0.0485, 0.1090, 0.4910, 0.0857, 0.0907, 0.1275, 0.0733], + device='cuda:0'), in_proj_covar=tensor([0.0508, 0.0418, 0.0419, 0.0523, 0.0413, 0.0424, 0.0409, 0.0366], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 21:05:07,012 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=139043.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:05:10,355 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=139048.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:05:15,387 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139055.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:05:18,013 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=139059.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:05:19,983 INFO [train.py:901] (0/4) Epoch 18, batch 1650, loss[loss=0.2028, simple_loss=0.2961, pruned_loss=0.0547, over 8037.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2944, pruned_loss=0.06696, over 1619252.82 frames. ], batch size: 22, lr: 4.26e-03, grad_scale: 8.0 +2023-02-06 21:05:28,835 INFO [optim.py:369] (0/4) 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 21:05:33,235 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139080.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:05:53,871 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1058, 1.6438, 3.2631, 1.3638, 2.1132, 3.5456, 3.6847, 3.0315], + device='cuda:0'), covar=tensor([0.0928, 0.1587, 0.0366, 0.2117, 0.1195, 0.0230, 0.0446, 0.0558], + device='cuda:0'), in_proj_covar=tensor([0.0286, 0.0317, 0.0279, 0.0310, 0.0300, 0.0257, 0.0398, 0.0302], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 21:05:54,423 INFO [train.py:901] (0/4) Epoch 18, batch 1700, loss[loss=0.2531, simple_loss=0.3345, pruned_loss=0.08586, over 8774.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.2943, pruned_loss=0.06745, over 1619085.86 frames. ], batch size: 30, lr: 4.26e-03, grad_scale: 8.0 +2023-02-06 21:06:28,654 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9803, 1.5666, 3.4153, 1.4690, 2.3191, 3.7365, 3.7879, 2.9866], + device='cuda:0'), covar=tensor([0.1109, 0.1833, 0.0413, 0.2192, 0.1217, 0.0261, 0.0586, 0.0726], + device='cuda:0'), in_proj_covar=tensor([0.0285, 0.0316, 0.0278, 0.0309, 0.0299, 0.0256, 0.0397, 0.0301], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-06 21:06:28,678 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139158.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:06:29,046 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 +2023-02-06 21:06:30,614 INFO [train.py:901] (0/4) Epoch 18, batch 1750, loss[loss=0.1975, simple_loss=0.2768, pruned_loss=0.05914, over 7691.00 frames. ], tot_loss[loss=0.2136, simple_loss=0.2934, pruned_loss=0.06691, over 1619506.69 frames. ], batch size: 18, lr: 4.25e-03, grad_scale: 8.0 +2023-02-06 21:06:32,192 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139163.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:06:39,591 INFO [optim.py:369] (0/4) 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 21:06:39,803 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 18, batch 1800, loss[loss=0.2186, simple_loss=0.3005, pruned_loss=0.06836, over 8246.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2929, pruned_loss=0.06681, over 1617497.32 frames. ], batch size: 24, lr: 4.25e-03, grad_scale: 8.0 +2023-02-06 21:07:37,620 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 18, batch 1850, loss[loss=0.244, simple_loss=0.3177, pruned_loss=0.08519, over 8330.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2931, pruned_loss=0.06686, over 1616584.99 frames. ], batch size: 25, lr: 4.25e-03, grad_scale: 8.0 +2023-02-06 21:07:49,460 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=139271.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:07:51,404 INFO [optim.py:369] (0/4) 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 21:08:03,886 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4857, 1.4462, 2.3940, 1.1887, 2.1603, 2.5733, 2.6861, 2.1921], + device='cuda:0'), covar=tensor([0.0995, 0.1225, 0.0445, 0.2087, 0.0704, 0.0363, 0.0659, 0.0717], + device='cuda:0'), in_proj_covar=tensor([0.0281, 0.0311, 0.0274, 0.0305, 0.0294, 0.0253, 0.0392, 0.0297], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-06 21:08:05,260 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139294.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:08:14,579 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139307.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:08:17,149 INFO [train.py:901] (0/4) Epoch 18, batch 1900, loss[loss=0.2074, simple_loss=0.2912, pruned_loss=0.06182, over 8234.00 frames. ], tot_loss[loss=0.2137, simple_loss=0.2932, pruned_loss=0.06706, over 1609624.80 frames. ], batch size: 22, lr: 4.25e-03, grad_scale: 8.0 +2023-02-06 21:08:52,414 INFO [train.py:901] (0/4) Epoch 18, batch 1950, loss[loss=0.1754, simple_loss=0.2573, pruned_loss=0.04677, over 7436.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2931, pruned_loss=0.06692, over 1611858.20 frames. ], batch size: 17, lr: 4.25e-03, grad_scale: 8.0 +2023-02-06 21:08:55,250 WARNING [train.py:1067] (0/4) 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] (0/4) 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 21:09:08,116 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-06 21:09:11,151 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139386.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:09:28,242 INFO [train.py:901] (0/4) Epoch 18, batch 2000, loss[loss=0.2267, simple_loss=0.3131, pruned_loss=0.07018, over 8718.00 frames. ], tot_loss[loss=0.2139, simple_loss=0.2935, pruned_loss=0.06712, over 1613983.50 frames. ], batch size: 34, lr: 4.25e-03, grad_scale: 8.0 +2023-02-06 21:09:28,248 WARNING [train.py:1067] (0/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-06 21:09:30,562 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139414.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:09:33,877 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139419.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:09:42,126 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139430.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:09:48,231 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139439.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:09:51,625 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139444.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:09:59,139 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139455.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:10:02,969 INFO [train.py:901] (0/4) Epoch 18, batch 2050, loss[loss=0.2514, simple_loss=0.3341, pruned_loss=0.08429, over 8358.00 frames. ], tot_loss[loss=0.214, simple_loss=0.2936, pruned_loss=0.06717, over 1620551.58 frames. ], batch size: 24, lr: 4.25e-03, grad_scale: 8.0 +2023-02-06 21:10:12,670 INFO [optim.py:369] (0/4) 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 21:10:39,806 INFO [train.py:901] (0/4) Epoch 18, batch 2100, loss[loss=0.3098, simple_loss=0.3707, pruned_loss=0.1244, over 8808.00 frames. ], tot_loss[loss=0.2136, simple_loss=0.2936, pruned_loss=0.06687, over 1619282.74 frames. ], batch size: 40, lr: 4.25e-03, grad_scale: 8.0 +2023-02-06 21:10:59,676 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 +2023-02-06 21:11:15,317 INFO [train.py:901] (0/4) Epoch 18, batch 2150, loss[loss=0.1793, simple_loss=0.2654, pruned_loss=0.04657, over 7805.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.2936, pruned_loss=0.06641, over 1622683.96 frames. ], batch size: 20, lr: 4.25e-03, grad_scale: 8.0 +2023-02-06 21:11:24,958 INFO [optim.py:369] (0/4) 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 21:11:34,704 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139589.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:11:43,096 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=139600.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:11:43,169 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 18, batch 2200, loss[loss=0.2165, simple_loss=0.2965, pruned_loss=0.06824, over 7796.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.2928, pruned_loss=0.06612, over 1619908.84 frames. ], batch size: 20, lr: 4.25e-03, grad_scale: 8.0 +2023-02-06 21:11:54,351 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 +2023-02-06 21:12:06,586 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7451, 1.7105, 2.2314, 1.5377, 1.3133, 2.3052, 0.3569, 1.4421], + device='cuda:0'), covar=tensor([0.1627, 0.1456, 0.0402, 0.1334, 0.2861, 0.0459, 0.2404, 0.1515], + device='cuda:0'), in_proj_covar=tensor([0.0179, 0.0189, 0.0118, 0.0217, 0.0263, 0.0126, 0.0165, 0.0184], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 21:12:10,494 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=139638.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:12:13,212 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139642.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:12:19,357 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=139651.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:12:26,638 INFO [train.py:901] (0/4) Epoch 18, batch 2250, loss[loss=0.1962, simple_loss=0.2728, pruned_loss=0.05982, over 7421.00 frames. ], tot_loss[loss=0.2143, simple_loss=0.2944, pruned_loss=0.06707, over 1619966.11 frames. ], batch size: 17, lr: 4.25e-03, grad_scale: 8.0 +2023-02-06 21:12:31,140 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139667.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:12:36,177 INFO [optim.py:369] (0/4) 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 21:12:51,929 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7112, 1.3965, 1.6282, 1.3286, 0.9279, 1.4123, 1.5650, 1.6174], + device='cuda:0'), covar=tensor([0.0433, 0.1052, 0.1414, 0.1205, 0.0505, 0.1225, 0.0563, 0.0491], + device='cuda:0'), in_proj_covar=tensor([0.0098, 0.0153, 0.0191, 0.0159, 0.0100, 0.0162, 0.0114, 0.0139], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 21:13:01,643 INFO [train.py:901] (0/4) Epoch 18, batch 2300, loss[loss=0.2298, simple_loss=0.3145, pruned_loss=0.07257, over 8292.00 frames. ], tot_loss[loss=0.2144, simple_loss=0.2947, pruned_loss=0.06704, over 1616880.26 frames. ], batch size: 23, lr: 4.25e-03, grad_scale: 8.0 +2023-02-06 21:13:04,651 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139715.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:13:18,653 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-06 21:13:32,013 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139753.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:13:37,206 INFO [train.py:901] (0/4) Epoch 18, batch 2350, loss[loss=0.2004, simple_loss=0.2924, pruned_loss=0.05421, over 8118.00 frames. ], tot_loss[loss=0.2153, simple_loss=0.2955, pruned_loss=0.06758, over 1613008.46 frames. ], batch size: 22, lr: 4.25e-03, grad_scale: 8.0 +2023-02-06 21:13:40,617 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139766.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:13:47,220 INFO [optim.py:369] (0/4) 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 21:13:55,495 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139787.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:14:08,992 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139807.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:14:11,587 INFO [train.py:901] (0/4) Epoch 18, batch 2400, loss[loss=0.2289, simple_loss=0.3091, pruned_loss=0.0743, over 8472.00 frames. ], tot_loss[loss=0.215, simple_loss=0.2951, pruned_loss=0.06744, over 1613696.92 frames. ], batch size: 29, lr: 4.24e-03, grad_scale: 8.0 +2023-02-06 21:14:20,256 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139822.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:14:21,638 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139824.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:14:48,491 INFO [train.py:901] (0/4) Epoch 18, batch 2450, loss[loss=0.1804, simple_loss=0.2667, pruned_loss=0.0471, over 7431.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.2952, pruned_loss=0.06792, over 1608880.20 frames. ], batch size: 17, lr: 4.24e-03, grad_scale: 8.0 +2023-02-06 21:14:53,478 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139868.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:14:58,170 INFO [optim.py:369] (0/4) 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 21:15:01,670 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.4694, 4.3967, 4.0333, 2.1629, 3.9190, 4.1143, 4.0009, 3.8488], + device='cuda:0'), covar=tensor([0.0731, 0.0546, 0.0992, 0.4497, 0.0798, 0.0949, 0.1162, 0.0806], + device='cuda:0'), in_proj_covar=tensor([0.0506, 0.0414, 0.0416, 0.0517, 0.0410, 0.0420, 0.0405, 0.0364], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 21:15:16,425 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0871, 1.2881, 1.2047, 0.6252, 1.1880, 1.0139, 0.1355, 1.1983], + device='cuda:0'), covar=tensor([0.0331, 0.0324, 0.0293, 0.0494, 0.0369, 0.0865, 0.0677, 0.0280], + device='cuda:0'), in_proj_covar=tensor([0.0430, 0.0371, 0.0321, 0.0426, 0.0357, 0.0515, 0.0377, 0.0398], + device='cuda:0'), out_proj_covar=tensor([1.1723e-04, 9.8204e-05, 8.4958e-05, 1.1336e-04, 9.5169e-05, 1.4766e-04, + 1.0272e-04, 1.0667e-04], device='cuda:0') +2023-02-06 21:15:23,630 INFO [train.py:901] (0/4) Epoch 18, batch 2500, loss[loss=0.1784, simple_loss=0.2715, pruned_loss=0.04266, over 8123.00 frames. ], tot_loss[loss=0.2154, simple_loss=0.2957, pruned_loss=0.06752, over 1613329.94 frames. ], batch size: 22, lr: 4.24e-03, grad_scale: 8.0 +2023-02-06 21:15:39,676 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=139933.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:15:47,234 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=139944.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:16:00,238 INFO [train.py:901] (0/4) Epoch 18, batch 2550, loss[loss=0.2173, simple_loss=0.2926, pruned_loss=0.071, over 7938.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.2964, pruned_loss=0.06799, over 1617253.54 frames. ], batch size: 20, lr: 4.24e-03, grad_scale: 8.0 +2023-02-06 21:16:07,329 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139971.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:16:09,810 INFO [optim.py:369] (0/4) 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 21:16:25,079 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139996.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:16:27,849 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-140000.pt +2023-02-06 21:16:35,516 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140009.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:16:36,624 INFO [train.py:901] (0/4) Epoch 18, batch 2600, loss[loss=0.2276, simple_loss=0.3028, pruned_loss=0.07619, over 8236.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2958, pruned_loss=0.06792, over 1613467.91 frames. ], batch size: 22, lr: 4.24e-03, grad_scale: 8.0 +2023-02-06 21:16:44,591 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140022.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:16:52,699 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140034.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:17:01,552 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140047.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:17:02,210 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140048.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:17:10,521 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140059.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:17:11,786 INFO [train.py:901] (0/4) Epoch 18, batch 2650, loss[loss=0.2484, simple_loss=0.3197, pruned_loss=0.08853, over 8454.00 frames. ], tot_loss[loss=0.2176, simple_loss=0.2971, pruned_loss=0.0691, over 1610990.38 frames. ], batch size: 27, lr: 4.24e-03, grad_scale: 8.0 +2023-02-06 21:17:18,364 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3413, 2.5462, 2.2004, 3.7194, 1.5963, 1.7778, 2.2353, 2.8246], + device='cuda:0'), covar=tensor([0.0804, 0.0844, 0.1012, 0.0389, 0.1222, 0.1410, 0.1151, 0.0820], + device='cuda:0'), in_proj_covar=tensor([0.0237, 0.0202, 0.0254, 0.0215, 0.0209, 0.0253, 0.0258, 0.0214], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 21:17:22,343 INFO [optim.py:369] (0/4) 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 21:17:47,904 INFO [train.py:901] (0/4) Epoch 18, batch 2700, loss[loss=0.2502, simple_loss=0.3076, pruned_loss=0.09638, over 7632.00 frames. ], tot_loss[loss=0.2171, simple_loss=0.2968, pruned_loss=0.06867, over 1608148.48 frames. ], batch size: 19, lr: 4.24e-03, grad_scale: 8.0 +2023-02-06 21:17:50,477 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.70 vs. limit=2.0 +2023-02-06 21:17:55,817 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140121.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:18:00,547 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140128.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:18:02,534 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=140151.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:18:19,415 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 +2023-02-06 21:18:23,207 INFO [train.py:901] (0/4) Epoch 18, batch 2750, loss[loss=0.2148, simple_loss=0.3078, pruned_loss=0.0609, over 8367.00 frames. ], tot_loss[loss=0.2179, simple_loss=0.2977, pruned_loss=0.06899, over 1609848.61 frames. ], batch size: 24, lr: 4.24e-03, grad_scale: 8.0 +2023-02-06 21:18:26,679 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=140166.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:18:28,866 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=140168.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:18:33,789 INFO [optim.py:369] (0/4) 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 21:19:00,762 INFO [train.py:901] (0/4) Epoch 18, batch 2800, loss[loss=0.2478, simple_loss=0.3228, pruned_loss=0.08638, over 8557.00 frames. ], tot_loss[loss=0.2173, simple_loss=0.2969, pruned_loss=0.06886, over 1608745.03 frames. ], batch size: 31, lr: 4.24e-03, grad_scale: 8.0 +2023-02-06 21:19:01,488 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=140212.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:19:25,705 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140246.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:19:35,887 INFO [train.py:901] (0/4) Epoch 18, batch 2850, loss[loss=0.1827, simple_loss=0.2642, pruned_loss=0.05062, over 7232.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.2957, pruned_loss=0.06763, over 1613129.12 frames. ], batch size: 16, lr: 4.24e-03, grad_scale: 8.0 +2023-02-06 21:19:39,551 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140266.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:19:45,686 INFO [optim.py:369] (0/4) 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 21:19:47,265 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140277.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:19:50,879 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140281.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:19:52,281 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140283.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:20:07,401 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140304.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:20:11,946 INFO [train.py:901] (0/4) Epoch 18, batch 2900, loss[loss=0.1774, simple_loss=0.2532, pruned_loss=0.0508, over 7801.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.2953, pruned_loss=0.06751, over 1612445.53 frames. ], batch size: 19, lr: 4.24e-03, grad_scale: 8.0 +2023-02-06 21:20:14,969 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140315.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:20:18,626 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.66 vs. limit=2.0 +2023-02-06 21:20:23,105 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4080, 2.3762, 1.6932, 2.1021, 1.9708, 1.4802, 1.8787, 1.9334], + device='cuda:0'), covar=tensor([0.1366, 0.0388, 0.1231, 0.0576, 0.0746, 0.1522, 0.0945, 0.0822], + device='cuda:0'), in_proj_covar=tensor([0.0350, 0.0234, 0.0322, 0.0301, 0.0296, 0.0330, 0.0339, 0.0312], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 21:20:23,751 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140327.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:20:25,180 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140329.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:20:32,779 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140340.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:20:44,326 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-06 21:20:47,982 INFO [train.py:901] (0/4) Epoch 18, batch 2950, loss[loss=0.2831, simple_loss=0.3372, pruned_loss=0.1145, over 6505.00 frames. ], tot_loss[loss=0.2143, simple_loss=0.2945, pruned_loss=0.06705, over 1607118.59 frames. ], batch size: 71, lr: 4.24e-03, grad_scale: 8.0 +2023-02-06 21:20:57,358 INFO [optim.py:369] (0/4) 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 21:21:18,961 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 18, batch 3000, loss[loss=0.2111, simple_loss=0.2947, pruned_loss=0.06373, over 8185.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.2951, pruned_loss=0.0672, over 1610843.01 frames. ], batch size: 23, lr: 4.24e-03, grad_scale: 8.0 +2023-02-06 21:21:23,798 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 21:21:37,685 INFO [train.py:935] (0/4) Epoch 18, validation: loss=0.1773, simple_loss=0.2774, pruned_loss=0.03861, over 944034.00 frames. +2023-02-06 21:21:37,686 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-06 21:22:03,257 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 +2023-02-06 21:22:07,939 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0480, 1.7853, 2.3657, 1.9588, 2.2423, 2.0382, 1.7573, 1.1117], + device='cuda:0'), covar=tensor([0.5316, 0.4620, 0.1855, 0.3342, 0.2374, 0.2698, 0.1889, 0.4813], + device='cuda:0'), in_proj_covar=tensor([0.0928, 0.0946, 0.0784, 0.0911, 0.0979, 0.0867, 0.0730, 0.0807], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 21:22:14,081 INFO [train.py:901] (0/4) Epoch 18, batch 3050, loss[loss=0.1762, simple_loss=0.2575, pruned_loss=0.04747, over 7696.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.2951, pruned_loss=0.06701, over 1613444.38 frames. ], batch size: 18, lr: 4.23e-03, grad_scale: 8.0 +2023-02-06 21:22:16,898 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=140465.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:22:21,637 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=140472.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:22:24,217 INFO [optim.py:369] (0/4) 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 21:22:42,883 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140502.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:22:46,233 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140507.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:22:48,872 INFO [train.py:901] (0/4) Epoch 18, batch 3100, loss[loss=0.1935, simple_loss=0.2665, pruned_loss=0.06031, over 7802.00 frames. ], tot_loss[loss=0.2153, simple_loss=0.2956, pruned_loss=0.06745, over 1613361.62 frames. ], batch size: 19, lr: 4.23e-03, grad_scale: 8.0 +2023-02-06 21:22:56,952 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140522.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:23:00,946 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140527.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:23:01,005 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140527.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:23:07,881 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140537.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:23:09,280 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140547.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:23:25,508 INFO [train.py:901] (0/4) Epoch 18, batch 3150, loss[loss=0.2321, simple_loss=0.3103, pruned_loss=0.07694, over 8113.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.2957, pruned_loss=0.06777, over 1614437.85 frames. ], batch size: 23, lr: 4.23e-03, grad_scale: 8.0 +2023-02-06 21:23:26,378 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140562.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:23:27,674 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140564.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:23:28,414 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6328, 2.0080, 3.3599, 1.4097, 2.5599, 2.1128, 1.6888, 2.5374], + device='cuda:0'), covar=tensor([0.1780, 0.2593, 0.0705, 0.4401, 0.1611, 0.2905, 0.2156, 0.2101], + device='cuda:0'), in_proj_covar=tensor([0.0510, 0.0573, 0.0549, 0.0618, 0.0632, 0.0575, 0.0511, 0.0623], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 21:23:34,933 INFO [optim.py:369] (0/4) 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] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140580.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:23:40,631 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140583.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:23:43,408 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140587.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:23:59,142 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140608.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:24:01,015 INFO [train.py:901] (0/4) Epoch 18, batch 3200, loss[loss=0.2266, simple_loss=0.308, pruned_loss=0.07259, over 8703.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.2961, pruned_loss=0.06766, over 1619043.04 frames. ], batch size: 49, lr: 4.23e-03, grad_scale: 8.0 +2023-02-06 21:24:07,874 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=140621.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:24:36,348 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4953, 1.8273, 1.8992, 1.1412, 1.9526, 1.4058, 0.3910, 1.7348], + device='cuda:0'), covar=tensor([0.0476, 0.0316, 0.0253, 0.0478, 0.0342, 0.0793, 0.0754, 0.0228], + device='cuda:0'), in_proj_covar=tensor([0.0436, 0.0376, 0.0322, 0.0431, 0.0361, 0.0522, 0.0380, 0.0400], + device='cuda:0'), out_proj_covar=tensor([1.1874e-04, 9.9782e-05, 8.5328e-05, 1.1489e-04, 9.6339e-05, 1.4965e-04, + 1.0343e-04, 1.0693e-04], device='cuda:0') +2023-02-06 21:24:36,812 INFO [train.py:901] (0/4) Epoch 18, batch 3250, loss[loss=0.195, simple_loss=0.2807, pruned_loss=0.05469, over 7822.00 frames. ], tot_loss[loss=0.2139, simple_loss=0.2949, pruned_loss=0.06641, over 1619307.22 frames. ], batch size: 20, lr: 4.23e-03, grad_scale: 8.0 +2023-02-06 21:24:40,492 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5867, 1.8676, 1.9466, 1.1748, 1.9850, 1.4833, 0.4350, 1.8236], + device='cuda:0'), covar=tensor([0.0428, 0.0325, 0.0231, 0.0445, 0.0357, 0.0796, 0.0703, 0.0202], + device='cuda:0'), in_proj_covar=tensor([0.0437, 0.0376, 0.0323, 0.0432, 0.0362, 0.0522, 0.0380, 0.0400], + device='cuda:0'), out_proj_covar=tensor([1.1882e-04, 9.9746e-05, 8.5410e-05, 1.1498e-04, 9.6435e-05, 1.4972e-04, + 1.0347e-04, 1.0694e-04], device='cuda:0') +2023-02-06 21:24:46,448 INFO [optim.py:369] (0/4) 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 21:25:13,073 INFO [train.py:901] (0/4) Epoch 18, batch 3300, loss[loss=0.2653, simple_loss=0.3401, pruned_loss=0.09526, over 8475.00 frames. ], tot_loss[loss=0.213, simple_loss=0.2937, pruned_loss=0.06613, over 1616809.68 frames. ], batch size: 25, lr: 4.23e-03, grad_scale: 8.0 +2023-02-06 21:25:30,579 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140741.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:25:35,580 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 +2023-02-06 21:25:39,312 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=140749.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:25:47,446 INFO [train.py:901] (0/4) Epoch 18, batch 3350, loss[loss=0.1819, simple_loss=0.2681, pruned_loss=0.04785, over 8109.00 frames. ], tot_loss[loss=0.2136, simple_loss=0.294, pruned_loss=0.0666, over 1618807.82 frames. ], batch size: 21, lr: 4.23e-03, grad_scale: 8.0 +2023-02-06 21:25:56,405 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.8193, 2.0346, 5.9450, 2.1350, 5.2941, 4.9915, 5.5097, 5.3866], + device='cuda:0'), covar=tensor([0.0432, 0.4027, 0.0306, 0.3551, 0.0944, 0.0808, 0.0460, 0.0449], + device='cuda:0'), in_proj_covar=tensor([0.0585, 0.0616, 0.0663, 0.0593, 0.0671, 0.0574, 0.0571, 0.0639], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 21:25:57,593 INFO [optim.py:369] (0/4) 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 21:26:23,945 INFO [train.py:901] (0/4) Epoch 18, batch 3400, loss[loss=0.2025, simple_loss=0.2888, pruned_loss=0.05807, over 8326.00 frames. ], tot_loss[loss=0.2139, simple_loss=0.2944, pruned_loss=0.06667, over 1618588.97 frames. ], batch size: 26, lr: 4.23e-03, grad_scale: 8.0 +2023-02-06 21:26:35,788 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140827.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:26:42,130 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140836.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:26:46,772 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140843.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:26:52,063 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=140851.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:26:58,804 INFO [train.py:901] (0/4) Epoch 18, batch 3450, loss[loss=0.2141, simple_loss=0.289, pruned_loss=0.06963, over 7420.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.2956, pruned_loss=0.0674, over 1616405.81 frames. ], batch size: 17, lr: 4.23e-03, grad_scale: 8.0 +2023-02-06 21:26:59,031 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140861.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:27:01,050 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140864.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:27:03,738 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=140871.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:27:08,275 INFO [optim.py:369] (0/4) 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 21:27:34,152 INFO [train.py:901] (0/4) Epoch 18, batch 3500, loss[loss=0.2293, simple_loss=0.314, pruned_loss=0.07231, over 8179.00 frames. ], tot_loss[loss=0.2153, simple_loss=0.2953, pruned_loss=0.06761, over 1613700.00 frames. ], batch size: 23, lr: 4.23e-03, grad_scale: 8.0 +2023-02-06 21:27:51,063 WARNING [train.py:1067] (0/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-06 21:27:51,193 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140935.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:28:09,211 INFO [train.py:901] (0/4) Epoch 18, batch 3550, loss[loss=0.2617, simple_loss=0.3214, pruned_loss=0.101, over 7957.00 frames. ], tot_loss[loss=0.2144, simple_loss=0.2949, pruned_loss=0.06698, over 1615264.09 frames. ], batch size: 21, lr: 4.23e-03, grad_scale: 8.0 +2023-02-06 21:28:11,999 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140965.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:28:12,771 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140966.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:28:18,752 INFO [optim.py:369] (0/4) 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 21:28:26,415 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140986.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:28:30,666 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140992.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:28:44,283 INFO [train.py:901] (0/4) Epoch 18, batch 3600, loss[loss=0.2134, simple_loss=0.2918, pruned_loss=0.06751, over 7523.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.294, pruned_loss=0.06631, over 1616238.59 frames. ], batch size: 18, lr: 4.23e-03, grad_scale: 8.0 +2023-02-06 21:28:49,255 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141017.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:29:20,152 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 +2023-02-06 21:29:20,372 INFO [train.py:901] (0/4) Epoch 18, batch 3650, loss[loss=0.2072, simple_loss=0.2777, pruned_loss=0.0683, over 7795.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.2931, pruned_loss=0.06597, over 1615084.69 frames. ], batch size: 19, lr: 4.23e-03, grad_scale: 8.0 +2023-02-06 21:29:30,817 INFO [optim.py:369] (0/4) 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 21:29:37,715 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=141085.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:29:53,089 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.5960, 4.5951, 4.0987, 1.9303, 3.9773, 4.1831, 4.1562, 3.8660], + device='cuda:0'), covar=tensor([0.0765, 0.0569, 0.1106, 0.4713, 0.0963, 0.0948, 0.1284, 0.0898], + device='cuda:0'), in_proj_covar=tensor([0.0507, 0.0421, 0.0420, 0.0521, 0.0411, 0.0421, 0.0406, 0.0368], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 21:29:55,734 INFO [train.py:901] (0/4) Epoch 18, batch 3700, loss[loss=0.1617, simple_loss=0.2539, pruned_loss=0.03479, over 7931.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2931, pruned_loss=0.06619, over 1613042.12 frames. ], batch size: 20, lr: 4.23e-03, grad_scale: 8.0 +2023-02-06 21:29:57,132 WARNING [train.py:1067] (0/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-06 21:30:02,958 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141120.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:30:20,677 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141145.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:30:32,108 INFO [train.py:901] (0/4) Epoch 18, batch 3750, loss[loss=0.2001, simple_loss=0.2871, pruned_loss=0.05655, over 8243.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.2937, pruned_loss=0.06632, over 1613950.47 frames. ], batch size: 24, lr: 4.22e-03, grad_scale: 8.0 +2023-02-06 21:30:32,263 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141161.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:30:39,100 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 21:30:48,315 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1273, 1.7081, 3.4691, 1.4935, 2.4214, 3.8464, 3.8414, 3.3364], + device='cuda:0'), covar=tensor([0.1020, 0.1646, 0.0310, 0.1988, 0.1033, 0.0200, 0.0556, 0.0528], + device='cuda:0'), in_proj_covar=tensor([0.0282, 0.0310, 0.0273, 0.0304, 0.0296, 0.0253, 0.0395, 0.0295], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-06 21:31:00,278 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 18, batch 3800, loss[loss=0.2028, simple_loss=0.2657, pruned_loss=0.06998, over 7215.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2931, pruned_loss=0.06614, over 1610838.05 frames. ], batch size: 16, lr: 4.22e-03, grad_scale: 8.0 +2023-02-06 21:31:15,001 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141222.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:31:29,257 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141242.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:31:32,640 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141247.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:31:34,649 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.6694, 2.7082, 2.5779, 4.0535, 1.7225, 2.3036, 2.3688, 3.0439], + device='cuda:0'), covar=tensor([0.0668, 0.0826, 0.0752, 0.0220, 0.1117, 0.1152, 0.0983, 0.0722], + device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0201, 0.0251, 0.0213, 0.0207, 0.0250, 0.0255, 0.0212], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 21:31:42,608 INFO [train.py:901] (0/4) Epoch 18, batch 3850, loss[loss=0.2053, simple_loss=0.2986, pruned_loss=0.05601, over 7807.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2925, pruned_loss=0.06571, over 1612141.29 frames. ], batch size: 20, lr: 4.22e-03, grad_scale: 8.0 +2023-02-06 21:31:46,905 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141267.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:31:52,716 INFO [optim.py:369] (0/4) 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 21:31:53,730 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.59 vs. limit=2.0 +2023-02-06 21:31:55,475 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=141279.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:32:00,526 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141286.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:32:03,907 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-06 21:32:17,061 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=141309.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:32:18,354 INFO [train.py:901] (0/4) Epoch 18, batch 3900, loss[loss=0.1999, simple_loss=0.2857, pruned_loss=0.05704, over 8355.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2943, pruned_loss=0.06638, over 1615818.10 frames. ], batch size: 24, lr: 4.22e-03, grad_scale: 8.0 +2023-02-06 21:32:42,392 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141347.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:32:52,481 INFO [train.py:901] (0/4) Epoch 18, batch 3950, loss[loss=0.2078, simple_loss=0.2709, pruned_loss=0.07231, over 7806.00 frames. ], tot_loss[loss=0.2139, simple_loss=0.295, pruned_loss=0.06646, over 1621908.53 frames. ], batch size: 19, lr: 4.22e-03, grad_scale: 8.0 +2023-02-06 21:33:02,714 INFO [optim.py:369] (0/4) 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 21:33:15,874 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141394.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:33:27,609 INFO [train.py:901] (0/4) Epoch 18, batch 4000, loss[loss=0.2497, simple_loss=0.328, pruned_loss=0.08567, over 8678.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.2949, pruned_loss=0.067, over 1619036.66 frames. ], batch size: 39, lr: 4.22e-03, grad_scale: 16.0 +2023-02-06 21:33:34,272 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 +2023-02-06 21:33:37,280 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141424.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:33:44,486 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141435.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 21:33:58,723 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141456.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:34:00,785 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141459.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:34:02,050 INFO [train.py:901] (0/4) Epoch 18, batch 4050, loss[loss=0.2499, simple_loss=0.3185, pruned_loss=0.09063, over 8096.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2948, pruned_loss=0.06677, over 1620811.81 frames. ], batch size: 23, lr: 4.22e-03, grad_scale: 16.0 +2023-02-06 21:34:12,708 INFO [optim.py:369] (0/4) 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 21:34:16,924 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141481.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:34:34,612 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=141505.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:34:38,484 INFO [train.py:901] (0/4) Epoch 18, batch 4100, loss[loss=0.1995, simple_loss=0.2784, pruned_loss=0.06031, over 7925.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2951, pruned_loss=0.06666, over 1622023.80 frames. ], batch size: 20, lr: 4.22e-03, grad_scale: 16.0 +2023-02-06 21:35:00,216 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141542.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:35:13,072 INFO [train.py:901] (0/4) Epoch 18, batch 4150, loss[loss=0.202, simple_loss=0.2795, pruned_loss=0.0623, over 7917.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2944, pruned_loss=0.06628, over 1618764.98 frames. ], batch size: 20, lr: 4.22e-03, grad_scale: 16.0 +2023-02-06 21:35:17,348 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141567.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:35:21,556 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4589, 1.8089, 1.8464, 1.2013, 1.9665, 1.3822, 0.4262, 1.7339], + device='cuda:0'), covar=tensor([0.0467, 0.0279, 0.0239, 0.0474, 0.0340, 0.0822, 0.0738, 0.0227], + device='cuda:0'), in_proj_covar=tensor([0.0431, 0.0371, 0.0318, 0.0429, 0.0358, 0.0517, 0.0375, 0.0396], + device='cuda:0'), out_proj_covar=tensor([1.1729e-04, 9.8246e-05, 8.4250e-05, 1.1435e-04, 9.5413e-05, 1.4836e-04, + 1.0212e-04, 1.0581e-04], device='cuda:0') +2023-02-06 21:35:22,679 INFO [optim.py:369] (0/4) 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 21:35:37,186 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.53 vs. limit=2.0 +2023-02-06 21:35:48,909 INFO [train.py:901] (0/4) Epoch 18, batch 4200, loss[loss=0.204, simple_loss=0.2765, pruned_loss=0.06571, over 7683.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2939, pruned_loss=0.06592, over 1621292.01 frames. ], batch size: 18, lr: 4.22e-03, grad_scale: 16.0 +2023-02-06 21:35:49,763 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4427, 1.3905, 2.3567, 1.3638, 2.1856, 2.5406, 2.6957, 2.1822], + device='cuda:0'), covar=tensor([0.1084, 0.1241, 0.0449, 0.1910, 0.0675, 0.0389, 0.0680, 0.0683], + device='cuda:0'), in_proj_covar=tensor([0.0284, 0.0312, 0.0276, 0.0307, 0.0297, 0.0256, 0.0397, 0.0298], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-06 21:35:55,106 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141620.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:36:02,343 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-06 21:36:15,978 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141650.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:36:23,880 INFO [train.py:901] (0/4) Epoch 18, batch 4250, loss[loss=0.1814, simple_loss=0.2663, pruned_loss=0.04818, over 7544.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2931, pruned_loss=0.06536, over 1616106.47 frames. ], batch size: 18, lr: 4.22e-03, grad_scale: 16.0 +2023-02-06 21:36:24,606 WARNING [train.py:1067] (0/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-06 21:36:33,255 INFO [optim.py:369] (0/4) 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 21:36:33,452 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141675.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:36:36,884 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141680.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:36:43,992 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=141691.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:36:54,142 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141705.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:36:58,028 INFO [train.py:901] (0/4) Epoch 18, batch 4300, loss[loss=0.2243, simple_loss=0.309, pruned_loss=0.06981, over 8574.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.2945, pruned_loss=0.06595, over 1618831.43 frames. ], batch size: 34, lr: 4.22e-03, grad_scale: 16.0 +2023-02-06 21:37:32,921 INFO [train.py:901] (0/4) Epoch 18, batch 4350, loss[loss=0.1698, simple_loss=0.249, pruned_loss=0.04531, over 7793.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2943, pruned_loss=0.0663, over 1612847.26 frames. ], batch size: 19, lr: 4.22e-03, grad_scale: 16.0 +2023-02-06 21:37:43,198 INFO [optim.py:369] (0/4) 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 21:37:46,081 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=141779.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 21:37:54,207 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-06 21:38:02,431 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=141803.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:38:04,527 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141806.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:38:07,909 INFO [train.py:901] (0/4) Epoch 18, batch 4400, loss[loss=0.1965, simple_loss=0.2848, pruned_loss=0.05407, over 8312.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2944, pruned_loss=0.06703, over 1610825.20 frames. ], batch size: 25, lr: 4.21e-03, grad_scale: 16.0 +2023-02-06 21:38:36,619 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-06 21:38:43,852 INFO [train.py:901] (0/4) Epoch 18, batch 4450, loss[loss=0.2604, simple_loss=0.3296, pruned_loss=0.09559, over 6811.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2932, pruned_loss=0.06629, over 1611003.91 frames. ], batch size: 71, lr: 4.21e-03, grad_scale: 16.0 +2023-02-06 21:38:53,326 INFO [optim.py:369] (0/4) 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 21:38:54,252 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141876.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:39:07,096 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141894.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 21:39:11,812 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141901.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:39:18,280 INFO [train.py:901] (0/4) Epoch 18, batch 4500, loss[loss=0.1997, simple_loss=0.2951, pruned_loss=0.05213, over 8020.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2941, pruned_loss=0.06627, over 1619433.61 frames. ], batch size: 22, lr: 4.21e-03, grad_scale: 16.0 +2023-02-06 21:39:23,237 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141918.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:39:27,788 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-06 21:39:34,700 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141934.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:39:42,675 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141946.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:39:53,399 INFO [train.py:901] (0/4) Epoch 18, batch 4550, loss[loss=0.2025, simple_loss=0.2925, pruned_loss=0.0562, over 8445.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.2934, pruned_loss=0.06578, over 1616988.23 frames. ], batch size: 27, lr: 4.21e-03, grad_scale: 16.0 +2023-02-06 21:40:03,506 INFO [optim.py:369] (0/4) 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 21:40:20,442 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-142000.pt +2023-02-06 21:40:29,738 INFO [train.py:901] (0/4) Epoch 18, batch 4600, loss[loss=0.2138, simple_loss=0.2807, pruned_loss=0.07341, over 7654.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2935, pruned_loss=0.06594, over 1617080.33 frames. ], batch size: 19, lr: 4.21e-03, grad_scale: 16.0 +2023-02-06 21:41:04,189 INFO [train.py:901] (0/4) Epoch 18, batch 4650, loss[loss=0.1796, simple_loss=0.2645, pruned_loss=0.04733, over 8231.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2929, pruned_loss=0.06583, over 1614740.24 frames. ], batch size: 22, lr: 4.21e-03, grad_scale: 16.0 +2023-02-06 21:41:05,110 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=142062.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:41:13,898 INFO [optim.py:369] (0/4) 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 21:41:23,654 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=142087.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:41:39,047 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.5107, 2.2587, 3.2017, 2.4463, 2.9340, 2.5118, 2.1808, 1.6745], + device='cuda:0'), covar=tensor([0.4852, 0.5115, 0.1693, 0.3679, 0.2677, 0.2611, 0.1780, 0.5454], + device='cuda:0'), in_proj_covar=tensor([0.0925, 0.0945, 0.0781, 0.0908, 0.0979, 0.0861, 0.0729, 0.0808], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 21:41:39,476 INFO [train.py:901] (0/4) Epoch 18, batch 4700, loss[loss=0.1808, simple_loss=0.2741, pruned_loss=0.04378, over 8335.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2922, pruned_loss=0.06571, over 1615020.89 frames. ], batch size: 25, lr: 4.21e-03, grad_scale: 16.0 +2023-02-06 21:42:00,893 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.92 vs. limit=5.0 +2023-02-06 21:42:06,016 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=142150.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 21:42:13,362 INFO [train.py:901] (0/4) Epoch 18, batch 4750, loss[loss=0.1723, simple_loss=0.2573, pruned_loss=0.04368, over 8029.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.293, pruned_loss=0.06616, over 1611386.67 frames. ], batch size: 22, lr: 4.21e-03, grad_scale: 16.0 +2023-02-06 21:42:23,433 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=142174.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:42:23,890 INFO [optim.py:369] (0/4) 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 21:42:24,103 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=142175.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 21:42:30,579 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-06 21:42:32,637 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-06 21:42:38,957 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0848, 2.2753, 1.9843, 2.7805, 1.3625, 1.7701, 2.0655, 2.3826], + device='cuda:0'), covar=tensor([0.0690, 0.0765, 0.0912, 0.0368, 0.1082, 0.1236, 0.0861, 0.0685], + device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0200, 0.0251, 0.0213, 0.0206, 0.0249, 0.0256, 0.0212], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 21:42:41,665 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=142199.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:42:49,354 INFO [train.py:901] (0/4) Epoch 18, batch 4800, loss[loss=0.2442, simple_loss=0.3239, pruned_loss=0.08227, over 8247.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2934, pruned_loss=0.06666, over 1608512.01 frames. ], batch size: 24, lr: 4.21e-03, grad_scale: 16.0 +2023-02-06 21:43:23,849 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-06 21:43:24,502 INFO [train.py:901] (0/4) Epoch 18, batch 4850, loss[loss=0.2264, simple_loss=0.3091, pruned_loss=0.07187, over 8471.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.293, pruned_loss=0.06621, over 1610547.51 frames. ], batch size: 25, lr: 4.21e-03, grad_scale: 16.0 +2023-02-06 21:43:33,971 INFO [optim.py:369] (0/4) 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 21:43:36,085 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=142278.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:43:45,097 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=142290.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:43:59,058 INFO [train.py:901] (0/4) Epoch 18, batch 4900, loss[loss=0.2033, simple_loss=0.2995, pruned_loss=0.05357, over 8456.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.2929, pruned_loss=0.06597, over 1617395.54 frames. ], batch size: 25, lr: 4.21e-03, grad_scale: 16.0 +2023-02-06 21:44:34,495 INFO [train.py:901] (0/4) Epoch 18, batch 4950, loss[loss=0.2688, simple_loss=0.3554, pruned_loss=0.09108, over 8453.00 frames. ], tot_loss[loss=0.213, simple_loss=0.293, pruned_loss=0.06646, over 1610565.98 frames. ], batch size: 27, lr: 4.21e-03, grad_scale: 16.0 +2023-02-06 21:44:44,960 INFO [optim.py:369] (0/4) 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 21:44:57,234 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=142393.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:45:05,915 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=142405.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:45:09,747 INFO [train.py:901] (0/4) Epoch 18, batch 5000, loss[loss=0.3081, simple_loss=0.3673, pruned_loss=0.1244, over 6804.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2939, pruned_loss=0.06656, over 1609918.94 frames. ], batch size: 71, lr: 4.21e-03, grad_scale: 8.0 +2023-02-06 21:45:44,314 INFO [train.py:901] (0/4) Epoch 18, batch 5050, loss[loss=0.2249, simple_loss=0.3033, pruned_loss=0.07324, over 7802.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2919, pruned_loss=0.06599, over 1604298.69 frames. ], batch size: 19, lr: 4.21e-03, grad_scale: 8.0 +2023-02-06 21:45:46,570 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3484, 2.5358, 2.2285, 2.9776, 2.0993, 2.1917, 2.1721, 2.6495], + device='cuda:0'), covar=tensor([0.0592, 0.0616, 0.0711, 0.0465, 0.0808, 0.0960, 0.0729, 0.0572], + device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0200, 0.0252, 0.0213, 0.0206, 0.0249, 0.0256, 0.0212], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 21:45:54,481 INFO [optim.py:369] (0/4) 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 21:46:04,030 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-06 21:46:19,165 INFO [train.py:901] (0/4) Epoch 18, batch 5100, loss[loss=0.2205, simple_loss=0.3016, pruned_loss=0.06974, over 8236.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.2928, pruned_loss=0.06607, over 1604978.21 frames. ], batch size: 24, lr: 4.20e-03, grad_scale: 8.0 +2023-02-06 21:46:54,253 INFO [train.py:901] (0/4) Epoch 18, batch 5150, loss[loss=0.1883, simple_loss=0.2724, pruned_loss=0.05212, over 8493.00 frames. ], tot_loss[loss=0.2136, simple_loss=0.2937, pruned_loss=0.06673, over 1603706.70 frames. ], batch size: 26, lr: 4.20e-03, grad_scale: 8.0 +2023-02-06 21:47:04,406 INFO [optim.py:369] (0/4) 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 21:47:07,302 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.6687, 1.5310, 4.8609, 1.8683, 4.2766, 4.0662, 4.3711, 4.2314], + device='cuda:0'), covar=tensor([0.0542, 0.4422, 0.0462, 0.3831, 0.1081, 0.0923, 0.0569, 0.0606], + device='cuda:0'), in_proj_covar=tensor([0.0593, 0.0624, 0.0672, 0.0599, 0.0680, 0.0582, 0.0576, 0.0644], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 21:47:14,700 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=142591.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:47:15,673 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 +2023-02-06 21:47:29,007 INFO [train.py:901] (0/4) Epoch 18, batch 5200, loss[loss=0.2226, simple_loss=0.3014, pruned_loss=0.07192, over 8429.00 frames. ], tot_loss[loss=0.213, simple_loss=0.293, pruned_loss=0.06651, over 1604054.69 frames. ], batch size: 27, lr: 4.20e-03, grad_scale: 8.0 +2023-02-06 21:47:55,309 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=142649.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:48:02,980 INFO [train.py:901] (0/4) Epoch 18, batch 5250, loss[loss=0.2179, simple_loss=0.2985, pruned_loss=0.06865, over 8250.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.2936, pruned_loss=0.06633, over 1608161.75 frames. ], batch size: 24, lr: 4.20e-03, grad_scale: 8.0 +2023-02-06 21:48:03,209 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=142661.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:48:03,667 WARNING [train.py:1067] (0/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-06 21:48:13,207 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=142674.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:48:14,357 INFO [optim.py:369] (0/4) 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 21:48:21,222 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=142686.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:48:37,885 INFO [train.py:901] (0/4) Epoch 18, batch 5300, loss[loss=0.2848, simple_loss=0.3547, pruned_loss=0.1074, over 8635.00 frames. ], tot_loss[loss=0.214, simple_loss=0.2943, pruned_loss=0.06678, over 1608489.75 frames. ], batch size: 34, lr: 4.20e-03, grad_scale: 8.0 +2023-02-06 21:48:59,483 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=142742.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:49:12,889 INFO [train.py:901] (0/4) Epoch 18, batch 5350, loss[loss=0.2219, simple_loss=0.3001, pruned_loss=0.07183, over 8519.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2945, pruned_loss=0.06691, over 1609954.96 frames. ], batch size: 26, lr: 4.20e-03, grad_scale: 8.0 +2023-02-06 21:49:19,900 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 +2023-02-06 21:49:22,761 INFO [optim.py:369] (0/4) 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 21:49:30,454 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7473, 1.4044, 1.6510, 1.2104, 0.9040, 1.3778, 1.6021, 1.3169], + device='cuda:0'), covar=tensor([0.0589, 0.1354, 0.1741, 0.1545, 0.0614, 0.1594, 0.0730, 0.0708], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0152, 0.0190, 0.0158, 0.0100, 0.0161, 0.0113, 0.0139], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 21:49:48,112 INFO [train.py:901] (0/4) Epoch 18, batch 5400, loss[loss=0.1716, simple_loss=0.2565, pruned_loss=0.04334, over 7662.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2955, pruned_loss=0.06738, over 1616264.08 frames. ], batch size: 19, lr: 4.20e-03, grad_scale: 8.0 +2023-02-06 21:50:22,753 INFO [train.py:901] (0/4) Epoch 18, batch 5450, loss[loss=0.203, simple_loss=0.2866, pruned_loss=0.05976, over 8458.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2945, pruned_loss=0.06691, over 1613529.74 frames. ], batch size: 27, lr: 4.20e-03, grad_scale: 8.0 +2023-02-06 21:50:33,556 INFO [optim.py:369] (0/4) 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 21:50:50,000 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-06 21:50:51,468 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.5486, 4.5476, 4.0377, 2.0002, 3.9511, 4.1931, 4.0541, 3.9728], + device='cuda:0'), covar=tensor([0.0695, 0.0537, 0.1015, 0.4821, 0.0886, 0.0976, 0.1253, 0.0739], + device='cuda:0'), in_proj_covar=tensor([0.0504, 0.0418, 0.0416, 0.0515, 0.0409, 0.0416, 0.0400, 0.0365], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 21:50:58,889 INFO [train.py:901] (0/4) Epoch 18, batch 5500, loss[loss=0.2154, simple_loss=0.3069, pruned_loss=0.06193, over 8294.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2932, pruned_loss=0.06617, over 1617035.42 frames. ], batch size: 23, lr: 4.20e-03, grad_scale: 8.0 +2023-02-06 21:51:05,232 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-06 21:51:14,986 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=142935.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:51:18,577 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9922, 1.6725, 2.0858, 1.7836, 1.9716, 2.0080, 1.8163, 0.8089], + device='cuda:0'), covar=tensor([0.5312, 0.4494, 0.1767, 0.3262, 0.2370, 0.2820, 0.1962, 0.4756], + device='cuda:0'), in_proj_covar=tensor([0.0926, 0.0946, 0.0778, 0.0912, 0.0984, 0.0865, 0.0732, 0.0811], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 21:51:30,185 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.71 vs. limit=2.0 +2023-02-06 21:51:33,204 INFO [train.py:901] (0/4) Epoch 18, batch 5550, loss[loss=0.2449, simple_loss=0.3234, pruned_loss=0.08319, over 8468.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2925, pruned_loss=0.06641, over 1613002.67 frames. ], batch size: 29, lr: 4.20e-03, grad_scale: 8.0 +2023-02-06 21:51:43,334 INFO [optim.py:369] (0/4) 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 21:51:49,614 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7128, 2.3680, 1.8684, 2.1569, 2.0221, 1.7407, 1.9508, 2.1095], + device='cuda:0'), covar=tensor([0.0955, 0.0340, 0.0854, 0.0488, 0.0587, 0.1063, 0.0747, 0.0717], + device='cuda:0'), in_proj_covar=tensor([0.0352, 0.0236, 0.0325, 0.0305, 0.0297, 0.0330, 0.0343, 0.0316], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 21:52:02,287 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7016, 1.9367, 2.1143, 1.3939, 2.1437, 1.6071, 0.5785, 1.8486], + device='cuda:0'), covar=tensor([0.0524, 0.0355, 0.0246, 0.0439, 0.0362, 0.0724, 0.0752, 0.0239], + device='cuda:0'), in_proj_covar=tensor([0.0435, 0.0372, 0.0322, 0.0432, 0.0359, 0.0521, 0.0376, 0.0398], + device='cuda:0'), out_proj_covar=tensor([1.1821e-04, 9.8507e-05, 8.5268e-05, 1.1507e-04, 9.5506e-05, 1.4919e-04, + 1.0204e-04, 1.0606e-04], device='cuda:0') +2023-02-06 21:52:08,287 INFO [train.py:901] (0/4) Epoch 18, batch 5600, loss[loss=0.2105, simple_loss=0.2955, pruned_loss=0.06271, over 8501.00 frames. ], tot_loss[loss=0.2126, simple_loss=0.2925, pruned_loss=0.06634, over 1610496.99 frames. ], batch size: 26, lr: 4.20e-03, grad_scale: 8.0 +2023-02-06 21:52:36,274 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=143050.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:52:43,530 INFO [train.py:901] (0/4) Epoch 18, batch 5650, loss[loss=0.1957, simple_loss=0.2913, pruned_loss=0.0501, over 8500.00 frames. ], tot_loss[loss=0.213, simple_loss=0.2931, pruned_loss=0.06643, over 1612208.49 frames. ], batch size: 26, lr: 4.20e-03, grad_scale: 8.0 +2023-02-06 21:52:54,545 INFO [optim.py:369] (0/4) 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 21:53:00,419 WARNING [train.py:1067] (0/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-06 21:53:01,150 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=143086.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:53:18,752 INFO [train.py:901] (0/4) Epoch 18, batch 5700, loss[loss=0.1896, simple_loss=0.2746, pruned_loss=0.05228, over 7812.00 frames. ], tot_loss[loss=0.2126, simple_loss=0.2928, pruned_loss=0.06622, over 1611078.28 frames. ], batch size: 20, lr: 4.20e-03, grad_scale: 8.0 +2023-02-06 21:53:19,802 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.64 vs. limit=2.0 +2023-02-06 21:53:23,692 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2436, 2.0142, 2.7190, 2.2423, 2.6764, 2.2347, 1.9651, 1.5080], + device='cuda:0'), covar=tensor([0.5107, 0.4936, 0.1756, 0.3601, 0.2383, 0.2943, 0.1837, 0.5096], + device='cuda:0'), in_proj_covar=tensor([0.0919, 0.0941, 0.0775, 0.0908, 0.0974, 0.0860, 0.0726, 0.0807], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 21:53:53,726 INFO [train.py:901] (0/4) Epoch 18, batch 5750, loss[loss=0.2664, simple_loss=0.336, pruned_loss=0.09844, over 8603.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2924, pruned_loss=0.06602, over 1606139.84 frames. ], batch size: 39, lr: 4.19e-03, grad_scale: 8.0 +2023-02-06 21:54:04,019 INFO [optim.py:369] (0/4) 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 21:54:04,732 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-06 21:54:21,111 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.3788, 4.3288, 3.9129, 2.0407, 3.8252, 3.9197, 4.0124, 3.7582], + device='cuda:0'), covar=tensor([0.0901, 0.0657, 0.1184, 0.4702, 0.0980, 0.1297, 0.1292, 0.0956], + device='cuda:0'), in_proj_covar=tensor([0.0506, 0.0420, 0.0420, 0.0518, 0.0412, 0.0419, 0.0402, 0.0368], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 21:54:21,864 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=143201.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:54:28,668 INFO [train.py:901] (0/4) Epoch 18, batch 5800, loss[loss=0.239, simple_loss=0.3149, pruned_loss=0.08157, over 8247.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2926, pruned_loss=0.06586, over 1610917.08 frames. ], batch size: 24, lr: 4.19e-03, grad_scale: 8.0 +2023-02-06 21:55:04,396 INFO [train.py:901] (0/4) Epoch 18, batch 5850, loss[loss=0.224, simple_loss=0.3122, pruned_loss=0.06786, over 8324.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.2931, pruned_loss=0.06584, over 1609148.11 frames. ], batch size: 25, lr: 4.19e-03, grad_scale: 8.0 +2023-02-06 21:55:12,230 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0975, 2.3601, 2.6343, 1.6078, 2.6989, 1.7975, 1.5512, 2.0036], + device='cuda:0'), covar=tensor([0.0716, 0.0434, 0.0296, 0.0687, 0.0487, 0.0821, 0.0843, 0.0490], + device='cuda:0'), in_proj_covar=tensor([0.0436, 0.0371, 0.0321, 0.0431, 0.0360, 0.0521, 0.0378, 0.0398], + device='cuda:0'), out_proj_covar=tensor([1.1835e-04, 9.8228e-05, 8.4939e-05, 1.1489e-04, 9.5905e-05, 1.4910e-04, + 1.0265e-04, 1.0608e-04], device='cuda:0') +2023-02-06 21:55:15,561 INFO [optim.py:369] (0/4) 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 21:55:36,566 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=143306.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:55:39,811 INFO [train.py:901] (0/4) Epoch 18, batch 5900, loss[loss=0.2452, simple_loss=0.3207, pruned_loss=0.08484, over 7103.00 frames. ], tot_loss[loss=0.2107, simple_loss=0.2914, pruned_loss=0.06501, over 1604818.35 frames. ], batch size: 72, lr: 4.19e-03, grad_scale: 8.0 +2023-02-06 21:55:53,333 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=143331.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:55:59,091 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.48 vs. limit=5.0 +2023-02-06 21:56:02,435 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.5118, 3.0570, 4.6112, 2.2535, 3.6589, 2.9752, 2.7188, 3.2616], + device='cuda:0'), covar=tensor([0.1595, 0.2145, 0.0823, 0.4136, 0.1509, 0.2714, 0.1872, 0.2359], + device='cuda:0'), in_proj_covar=tensor([0.0510, 0.0573, 0.0547, 0.0618, 0.0632, 0.0578, 0.0511, 0.0624], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 21:56:14,454 INFO [train.py:901] (0/4) Epoch 18, batch 5950, loss[loss=0.2227, simple_loss=0.2994, pruned_loss=0.073, over 7925.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2904, pruned_loss=0.06466, over 1600664.64 frames. ], batch size: 20, lr: 4.19e-03, grad_scale: 8.0 +2023-02-06 21:56:25,148 INFO [optim.py:369] (0/4) 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 21:56:40,500 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 +2023-02-06 21:56:49,466 INFO [train.py:901] (0/4) Epoch 18, batch 6000, loss[loss=0.2261, simple_loss=0.3018, pruned_loss=0.07524, over 7646.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.29, pruned_loss=0.06475, over 1601314.22 frames. ], batch size: 19, lr: 4.19e-03, grad_scale: 8.0 +2023-02-06 21:56:49,468 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 21:57:03,429 INFO [train.py:935] (0/4) Epoch 18, validation: loss=0.1765, simple_loss=0.2767, pruned_loss=0.03814, over 944034.00 frames. +2023-02-06 21:57:03,430 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-06 21:57:08,530 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=143418.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 21:57:35,816 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=143457.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:57:38,438 INFO [train.py:901] (0/4) Epoch 18, batch 6050, loss[loss=0.1977, simple_loss=0.2613, pruned_loss=0.06703, over 7654.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2916, pruned_loss=0.06551, over 1602227.86 frames. ], batch size: 19, lr: 4.19e-03, grad_scale: 8.0 +2023-02-06 21:57:48,584 INFO [optim.py:369] (0/4) 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 21:57:49,089 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 +2023-02-06 21:57:52,900 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7581, 1.8953, 2.0815, 1.4847, 2.2484, 1.4144, 0.7315, 1.9326], + device='cuda:0'), covar=tensor([0.0561, 0.0347, 0.0295, 0.0512, 0.0368, 0.0949, 0.0773, 0.0300], + device='cuda:0'), in_proj_covar=tensor([0.0435, 0.0372, 0.0322, 0.0431, 0.0360, 0.0520, 0.0378, 0.0397], + device='cuda:0'), out_proj_covar=tensor([1.1818e-04, 9.8333e-05, 8.5246e-05, 1.1477e-04, 9.5715e-05, 1.4898e-04, + 1.0257e-04, 1.0579e-04], device='cuda:0') +2023-02-06 21:57:53,563 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=143482.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 21:57:59,582 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=143491.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 21:58:07,289 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7466, 1.8926, 1.6624, 2.3133, 1.0102, 1.4105, 1.6671, 1.8463], + device='cuda:0'), covar=tensor([0.0742, 0.0732, 0.0902, 0.0400, 0.1145, 0.1421, 0.0853, 0.0778], + device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0199, 0.0250, 0.0213, 0.0208, 0.0247, 0.0254, 0.0214], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 21:58:12,147 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.2485, 4.2117, 3.8686, 1.7709, 3.8038, 3.7972, 3.7483, 3.4931], + device='cuda:0'), covar=tensor([0.0742, 0.0580, 0.1027, 0.4893, 0.0939, 0.1045, 0.1430, 0.0778], + device='cuda:0'), in_proj_covar=tensor([0.0512, 0.0425, 0.0425, 0.0524, 0.0417, 0.0422, 0.0408, 0.0373], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 21:58:13,415 INFO [train.py:901] (0/4) Epoch 18, batch 6100, loss[loss=0.1987, simple_loss=0.2919, pruned_loss=0.05277, over 8184.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2909, pruned_loss=0.06546, over 1601026.56 frames. ], batch size: 23, lr: 4.19e-03, grad_scale: 8.0 +2023-02-06 21:58:39,437 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-06 21:58:49,846 INFO [train.py:901] (0/4) Epoch 18, batch 6150, loss[loss=0.2103, simple_loss=0.2992, pruned_loss=0.06072, over 8546.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2904, pruned_loss=0.06495, over 1605701.31 frames. ], batch size: 31, lr: 4.19e-03, grad_scale: 8.0 +2023-02-06 21:59:00,211 INFO [optim.py:369] (0/4) 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 21:59:02,802 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.44 vs. limit=5.0 +2023-02-06 21:59:25,630 INFO [train.py:901] (0/4) Epoch 18, batch 6200, loss[loss=0.2401, simple_loss=0.307, pruned_loss=0.08662, over 7153.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2922, pruned_loss=0.06584, over 1609386.61 frames. ], batch size: 72, lr: 4.19e-03, grad_scale: 8.0 +2023-02-06 21:59:26,581 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([0.9606, 1.0911, 1.0652, 0.5629, 1.0514, 0.9216, 0.0880, 1.0509], + device='cuda:0'), covar=tensor([0.0346, 0.0306, 0.0265, 0.0479, 0.0354, 0.0744, 0.0669, 0.0264], + device='cuda:0'), in_proj_covar=tensor([0.0439, 0.0375, 0.0325, 0.0437, 0.0364, 0.0526, 0.0382, 0.0401], + device='cuda:0'), out_proj_covar=tensor([1.1915e-04, 9.9282e-05, 8.6034e-05, 1.1630e-04, 9.7003e-05, 1.5043e-04, + 1.0363e-04, 1.0695e-04], device='cuda:0') +2023-02-06 22:00:01,315 INFO [train.py:901] (0/4) Epoch 18, batch 6250, loss[loss=0.2182, simple_loss=0.2978, pruned_loss=0.06931, over 7069.00 frames. ], tot_loss[loss=0.2114, simple_loss=0.2916, pruned_loss=0.06565, over 1607393.25 frames. ], batch size: 72, lr: 4.19e-03, grad_scale: 8.0 +2023-02-06 22:00:12,420 INFO [optim.py:369] (0/4) 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 22:00:28,278 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8438, 3.4906, 2.0812, 2.7089, 2.6827, 1.7959, 2.6536, 2.8579], + device='cuda:0'), covar=tensor([0.1882, 0.0460, 0.1322, 0.0901, 0.0904, 0.1699, 0.1208, 0.1294], + device='cuda:0'), in_proj_covar=tensor([0.0352, 0.0235, 0.0325, 0.0305, 0.0296, 0.0331, 0.0342, 0.0316], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 22:00:29,821 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.57 vs. limit=2.0 +2023-02-06 22:00:37,053 INFO [train.py:901] (0/4) Epoch 18, batch 6300, loss[loss=0.1909, simple_loss=0.284, pruned_loss=0.04886, over 8114.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2933, pruned_loss=0.06625, over 1609720.91 frames. ], batch size: 23, lr: 4.19e-03, grad_scale: 8.0 +2023-02-06 22:00:58,044 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.59 vs. limit=5.0 +2023-02-06 22:01:11,901 INFO [train.py:901] (0/4) Epoch 18, batch 6350, loss[loss=0.2102, simple_loss=0.2907, pruned_loss=0.06487, over 8459.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2926, pruned_loss=0.06557, over 1609462.66 frames. ], batch size: 29, lr: 4.19e-03, grad_scale: 8.0 +2023-02-06 22:01:13,340 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=143762.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 22:01:22,535 INFO [optim.py:369] (0/4) 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 22:01:40,208 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.29 vs. limit=5.0 +2023-02-06 22:01:47,357 INFO [train.py:901] (0/4) Epoch 18, batch 6400, loss[loss=0.21, simple_loss=0.2981, pruned_loss=0.06096, over 8488.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2939, pruned_loss=0.06647, over 1611776.37 frames. ], batch size: 29, lr: 4.19e-03, grad_scale: 8.0 +2023-02-06 22:01:54,087 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9342, 1.8246, 2.6412, 1.6806, 2.2078, 2.8935, 2.8432, 2.6196], + device='cuda:0'), covar=tensor([0.0911, 0.1267, 0.0518, 0.1606, 0.1210, 0.0264, 0.0808, 0.0431], + device='cuda:0'), in_proj_covar=tensor([0.0282, 0.0312, 0.0276, 0.0305, 0.0295, 0.0254, 0.0397, 0.0295], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-06 22:02:04,433 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=143835.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 22:02:21,896 INFO [train.py:901] (0/4) Epoch 18, batch 6450, loss[loss=0.2026, simple_loss=0.2795, pruned_loss=0.06288, over 8083.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2928, pruned_loss=0.06584, over 1613698.60 frames. ], batch size: 21, lr: 4.18e-03, grad_scale: 8.0 +2023-02-06 22:02:33,456 INFO [optim.py:369] (0/4) 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 22:02:34,287 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=143877.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 22:02:57,285 INFO [train.py:901] (0/4) Epoch 18, batch 6500, loss[loss=0.2028, simple_loss=0.2909, pruned_loss=0.05741, over 8330.00 frames. ], tot_loss[loss=0.2126, simple_loss=0.2932, pruned_loss=0.066, over 1614959.63 frames. ], batch size: 25, lr: 4.18e-03, grad_scale: 8.0 +2023-02-06 22:03:24,069 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 18, batch 6550, loss[loss=0.1889, simple_loss=0.2745, pruned_loss=0.05162, over 7977.00 frames. ], tot_loss[loss=0.2126, simple_loss=0.2934, pruned_loss=0.06591, over 1616023.45 frames. ], batch size: 21, lr: 4.18e-03, grad_scale: 8.0 +2023-02-06 22:03:41,856 INFO [optim.py:369] (0/4) 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 22:03:48,811 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-06 22:03:59,959 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-144000.pt +2023-02-06 22:04:08,860 INFO [train.py:901] (0/4) Epoch 18, batch 6600, loss[loss=0.206, simple_loss=0.2877, pruned_loss=0.06212, over 8596.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2927, pruned_loss=0.0658, over 1617123.17 frames. ], batch size: 34, lr: 4.18e-03, grad_scale: 8.0 +2023-02-06 22:04:10,894 WARNING [train.py:1067] (0/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-06 22:04:18,941 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 +2023-02-06 22:04:43,666 INFO [train.py:901] (0/4) Epoch 18, batch 6650, loss[loss=0.2349, simple_loss=0.3122, pruned_loss=0.07878, over 8283.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2922, pruned_loss=0.06559, over 1616262.18 frames. ], batch size: 23, lr: 4.18e-03, grad_scale: 8.0 +2023-02-06 22:04:54,721 INFO [optim.py:369] (0/4) 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 22:05:15,481 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 +2023-02-06 22:05:19,671 INFO [train.py:901] (0/4) Epoch 18, batch 6700, loss[loss=0.2385, simple_loss=0.3175, pruned_loss=0.0798, over 8316.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2928, pruned_loss=0.06583, over 1617017.77 frames. ], batch size: 25, lr: 4.18e-03, grad_scale: 8.0 +2023-02-06 22:05:34,393 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=144133.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 22:05:50,160 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 +2023-02-06 22:05:51,947 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=144158.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 22:05:53,840 INFO [train.py:901] (0/4) Epoch 18, batch 6750, loss[loss=0.1897, simple_loss=0.2941, pruned_loss=0.04271, over 8493.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2921, pruned_loss=0.06522, over 1610337.11 frames. ], batch size: 26, lr: 4.18e-03, grad_scale: 8.0 +2023-02-06 22:06:03,940 INFO [optim.py:369] (0/4) 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 22:06:18,014 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4216, 2.0864, 2.7406, 2.3550, 2.7919, 2.4129, 2.0990, 1.5667], + device='cuda:0'), covar=tensor([0.4987, 0.4715, 0.1662, 0.2965, 0.2010, 0.2549, 0.1733, 0.4606], + device='cuda:0'), in_proj_covar=tensor([0.0932, 0.0950, 0.0784, 0.0919, 0.0986, 0.0871, 0.0736, 0.0815], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 22:06:25,427 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=144206.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 22:06:27,270 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-06 22:06:28,559 INFO [train.py:901] (0/4) Epoch 18, batch 6800, loss[loss=0.2117, simple_loss=0.2792, pruned_loss=0.07214, over 8028.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2914, pruned_loss=0.0646, over 1611841.40 frames. ], batch size: 22, lr: 4.18e-03, grad_scale: 8.0 +2023-02-06 22:06:40,786 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7117, 1.9037, 1.6425, 2.2646, 1.0624, 1.4651, 1.5623, 1.8505], + device='cuda:0'), covar=tensor([0.0766, 0.0696, 0.0954, 0.0451, 0.1108, 0.1288, 0.0883, 0.0738], + device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0200, 0.0252, 0.0213, 0.0208, 0.0249, 0.0253, 0.0213], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 22:06:42,848 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=144231.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 22:07:03,970 INFO [train.py:901] (0/4) Epoch 18, batch 6850, loss[loss=0.2052, simple_loss=0.2673, pruned_loss=0.07157, over 7788.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2929, pruned_loss=0.0657, over 1610820.26 frames. ], batch size: 19, lr: 4.18e-03, grad_scale: 8.0 +2023-02-06 22:07:12,878 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6065, 1.3255, 1.6003, 1.2081, 0.8503, 1.3394, 1.4166, 1.3561], + device='cuda:0'), covar=tensor([0.0529, 0.1302, 0.1771, 0.1504, 0.0561, 0.1500, 0.0711, 0.0687], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0152, 0.0190, 0.0158, 0.0099, 0.0162, 0.0113, 0.0139], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 22:07:13,993 INFO [optim.py:369] (0/4) 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 22:07:17,332 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-06 22:07:18,819 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.8306, 1.4714, 3.9635, 1.4151, 3.5040, 3.2808, 3.6421, 3.4965], + device='cuda:0'), covar=tensor([0.0595, 0.4169, 0.0592, 0.3888, 0.1200, 0.1058, 0.0598, 0.0708], + device='cuda:0'), in_proj_covar=tensor([0.0592, 0.0626, 0.0668, 0.0599, 0.0677, 0.0582, 0.0576, 0.0646], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 22:07:38,061 INFO [train.py:901] (0/4) Epoch 18, batch 6900, loss[loss=0.2254, simple_loss=0.303, pruned_loss=0.07386, over 8505.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.2934, pruned_loss=0.06604, over 1614077.63 frames. ], batch size: 26, lr: 4.18e-03, grad_scale: 8.0 +2023-02-06 22:07:44,344 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-02-06 22:07:50,538 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 +2023-02-06 22:07:52,974 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3893, 2.6373, 3.0854, 1.9171, 3.2165, 2.0028, 1.5471, 2.2715], + device='cuda:0'), covar=tensor([0.0701, 0.0381, 0.0236, 0.0631, 0.0518, 0.0757, 0.0875, 0.0514], + device='cuda:0'), in_proj_covar=tensor([0.0438, 0.0376, 0.0324, 0.0435, 0.0365, 0.0525, 0.0381, 0.0403], + device='cuda:0'), out_proj_covar=tensor([1.1879e-04, 9.9485e-05, 8.5836e-05, 1.1576e-04, 9.7193e-05, 1.5025e-04, + 1.0338e-04, 1.0749e-04], device='cuda:0') +2023-02-06 22:08:13,484 INFO [train.py:901] (0/4) Epoch 18, batch 6950, loss[loss=0.228, simple_loss=0.3172, pruned_loss=0.06942, over 8470.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2948, pruned_loss=0.06677, over 1620300.44 frames. ], batch size: 29, lr: 4.18e-03, grad_scale: 8.0 +2023-02-06 22:08:24,082 INFO [optim.py:369] (0/4) 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 22:08:25,463 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-06 22:08:33,150 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4817, 1.9821, 2.9417, 1.4012, 2.1911, 1.7415, 1.7672, 2.1019], + device='cuda:0'), covar=tensor([0.2102, 0.2485, 0.0992, 0.4603, 0.2047, 0.3531, 0.2236, 0.2419], + device='cuda:0'), in_proj_covar=tensor([0.0514, 0.0580, 0.0551, 0.0623, 0.0640, 0.0585, 0.0516, 0.0629], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 22:08:47,782 INFO [train.py:901] (0/4) Epoch 18, batch 7000, loss[loss=0.1955, simple_loss=0.2886, pruned_loss=0.05123, over 8597.00 frames. ], tot_loss[loss=0.214, simple_loss=0.2949, pruned_loss=0.06652, over 1622842.64 frames. ], batch size: 34, lr: 4.18e-03, grad_scale: 16.0 +2023-02-06 22:09:01,088 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=144429.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:09:22,578 INFO [train.py:901] (0/4) Epoch 18, batch 7050, loss[loss=0.1986, simple_loss=0.284, pruned_loss=0.05662, over 7973.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.2937, pruned_loss=0.06623, over 1621222.81 frames. ], batch size: 21, lr: 4.18e-03, grad_scale: 16.0 +2023-02-06 22:09:34,230 INFO [optim.py:369] (0/4) 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 22:09:58,448 INFO [train.py:901] (0/4) Epoch 18, batch 7100, loss[loss=0.1728, simple_loss=0.2561, pruned_loss=0.04477, over 7246.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2934, pruned_loss=0.06618, over 1618342.16 frames. ], batch size: 16, lr: 4.18e-03, grad_scale: 16.0 +2023-02-06 22:10:33,601 INFO [train.py:901] (0/4) Epoch 18, batch 7150, loss[loss=0.1774, simple_loss=0.2687, pruned_loss=0.04307, over 8091.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2928, pruned_loss=0.06558, over 1615598.89 frames. ], batch size: 21, lr: 4.17e-03, grad_scale: 16.0 +2023-02-06 22:10:43,978 INFO [optim.py:369] (0/4) 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 22:10:53,082 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9662, 1.6551, 3.2753, 1.3769, 2.2166, 3.5552, 3.7069, 3.0135], + device='cuda:0'), covar=tensor([0.1119, 0.1601, 0.0347, 0.2138, 0.1092, 0.0224, 0.0521, 0.0618], + device='cuda:0'), in_proj_covar=tensor([0.0282, 0.0312, 0.0277, 0.0305, 0.0294, 0.0253, 0.0396, 0.0296], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-06 22:11:10,024 INFO [train.py:901] (0/4) Epoch 18, batch 7200, loss[loss=0.1641, simple_loss=0.2634, pruned_loss=0.03237, over 8252.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.294, pruned_loss=0.06573, over 1622389.02 frames. ], batch size: 24, lr: 4.17e-03, grad_scale: 16.0 +2023-02-06 22:11:44,447 INFO [train.py:901] (0/4) Epoch 18, batch 7250, loss[loss=0.2041, simple_loss=0.2773, pruned_loss=0.06546, over 7785.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2931, pruned_loss=0.06553, over 1614485.02 frames. ], batch size: 19, lr: 4.17e-03, grad_scale: 16.0 +2023-02-06 22:11:54,464 INFO [optim.py:369] (0/4) 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 22:12:19,771 INFO [train.py:901] (0/4) Epoch 18, batch 7300, loss[loss=0.2295, simple_loss=0.3162, pruned_loss=0.07134, over 8294.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2932, pruned_loss=0.06559, over 1616538.43 frames. ], batch size: 23, lr: 4.17e-03, grad_scale: 16.0 +2023-02-06 22:12:53,974 INFO [train.py:901] (0/4) Epoch 18, batch 7350, loss[loss=0.2235, simple_loss=0.307, pruned_loss=0.07003, over 8444.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2934, pruned_loss=0.06599, over 1616188.20 frames. ], batch size: 27, lr: 4.17e-03, grad_scale: 16.0 +2023-02-06 22:13:02,922 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=144773.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:13:04,751 INFO [optim.py:369] (0/4) 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 22:13:08,080 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-06 22:13:26,885 WARNING [train.py:1067] (0/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-06 22:13:28,902 INFO [train.py:901] (0/4) Epoch 18, batch 7400, loss[loss=0.1842, simple_loss=0.258, pruned_loss=0.05522, over 7532.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2927, pruned_loss=0.06572, over 1615430.96 frames. ], batch size: 18, lr: 4.17e-03, grad_scale: 16.0 +2023-02-06 22:14:04,315 INFO [train.py:901] (0/4) Epoch 18, batch 7450, loss[loss=0.1642, simple_loss=0.2446, pruned_loss=0.04187, over 7436.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.2934, pruned_loss=0.06638, over 1612834.19 frames. ], batch size: 17, lr: 4.17e-03, grad_scale: 16.0 +2023-02-06 22:14:07,807 WARNING [train.py:1067] (0/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-06 22:14:14,580 INFO [optim.py:369] (0/4) 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 22:14:22,651 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=144888.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:14:30,657 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2754, 2.5342, 2.2404, 3.4231, 1.7981, 1.8578, 2.1822, 2.7688], + device='cuda:0'), covar=tensor([0.0693, 0.0829, 0.0785, 0.0335, 0.1073, 0.1318, 0.1031, 0.0706], + device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0197, 0.0246, 0.0210, 0.0205, 0.0244, 0.0250, 0.0210], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 22:14:38,535 INFO [train.py:901] (0/4) Epoch 18, batch 7500, loss[loss=0.2471, simple_loss=0.3158, pruned_loss=0.08923, over 6840.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2938, pruned_loss=0.06647, over 1616256.09 frames. ], batch size: 71, lr: 4.17e-03, grad_scale: 8.0 +2023-02-06 22:15:12,904 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1649, 1.6731, 3.0806, 1.4200, 2.1761, 3.3023, 3.4878, 2.8655], + device='cuda:0'), covar=tensor([0.0964, 0.1599, 0.0413, 0.2176, 0.1073, 0.0274, 0.0645, 0.0621], + device='cuda:0'), in_proj_covar=tensor([0.0284, 0.0317, 0.0277, 0.0308, 0.0296, 0.0256, 0.0401, 0.0300], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-06 22:15:14,116 INFO [train.py:901] (0/4) Epoch 18, batch 7550, loss[loss=0.1901, simple_loss=0.272, pruned_loss=0.05409, over 7648.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2938, pruned_loss=0.06637, over 1618687.30 frames. ], batch size: 19, lr: 4.17e-03, grad_scale: 8.0 +2023-02-06 22:15:24,760 INFO [optim.py:369] (0/4) 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 22:15:48,836 INFO [train.py:901] (0/4) Epoch 18, batch 7600, loss[loss=0.2219, simple_loss=0.3097, pruned_loss=0.06699, over 8451.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.293, pruned_loss=0.06597, over 1619580.34 frames. ], batch size: 27, lr: 4.17e-03, grad_scale: 8.0 +2023-02-06 22:15:51,061 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145014.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:16:12,824 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145045.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:16:24,419 INFO [train.py:901] (0/4) Epoch 18, batch 7650, loss[loss=0.208, simple_loss=0.2978, pruned_loss=0.0591, over 8496.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2927, pruned_loss=0.06596, over 1614569.43 frames. ], batch size: 28, lr: 4.17e-03, grad_scale: 8.0 +2023-02-06 22:16:35,685 INFO [optim.py:369] (0/4) 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 22:16:40,588 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145084.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:16:58,562 INFO [train.py:901] (0/4) Epoch 18, batch 7700, loss[loss=0.2559, simple_loss=0.3297, pruned_loss=0.09107, over 8503.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.293, pruned_loss=0.066, over 1615423.05 frames. ], batch size: 28, lr: 4.17e-03, grad_scale: 8.0 +2023-02-06 22:17:16,306 WARNING [train.py:1067] (0/4) 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] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145144.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:17:33,670 INFO [train.py:901] (0/4) Epoch 18, batch 7750, loss[loss=0.2452, simple_loss=0.3231, pruned_loss=0.08368, over 8488.00 frames. ], tot_loss[loss=0.2126, simple_loss=0.293, pruned_loss=0.06607, over 1614976.29 frames. ], batch size: 28, lr: 4.17e-03, grad_scale: 8.0 +2023-02-06 22:17:40,033 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145169.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:17:44,020 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1993, 1.8477, 3.5817, 1.8429, 2.6992, 3.9169, 4.0088, 3.3941], + device='cuda:0'), covar=tensor([0.1087, 0.1542, 0.0317, 0.1821, 0.0923, 0.0225, 0.0500, 0.0554], + device='cuda:0'), in_proj_covar=tensor([0.0283, 0.0314, 0.0276, 0.0306, 0.0294, 0.0255, 0.0399, 0.0297], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-06 22:17:45,193 INFO [optim.py:369] (0/4) 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 22:18:08,794 INFO [train.py:901] (0/4) Epoch 18, batch 7800, loss[loss=0.2339, simple_loss=0.3178, pruned_loss=0.07497, over 8471.00 frames. ], tot_loss[loss=0.2137, simple_loss=0.2938, pruned_loss=0.06684, over 1615291.01 frames. ], batch size: 25, lr: 4.17e-03, grad_scale: 8.0 +2023-02-06 22:18:30,140 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7059, 1.6904, 2.2930, 1.4841, 1.2208, 2.1947, 0.3956, 1.3801], + device='cuda:0'), covar=tensor([0.1750, 0.1290, 0.0353, 0.1471, 0.2996, 0.0487, 0.2377, 0.1375], + device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0190, 0.0121, 0.0215, 0.0266, 0.0129, 0.0165, 0.0182], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 22:18:36,389 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7529, 1.7066, 2.3488, 1.6355, 1.3733, 2.2812, 0.4075, 1.5113], + device='cuda:0'), covar=tensor([0.1828, 0.1261, 0.0330, 0.1364, 0.2794, 0.0444, 0.2409, 0.1357], + device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0190, 0.0121, 0.0215, 0.0266, 0.0129, 0.0165, 0.0182], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 22:18:42,840 INFO [train.py:901] (0/4) Epoch 18, batch 7850, loss[loss=0.1791, simple_loss=0.2727, pruned_loss=0.04281, over 8024.00 frames. ], tot_loss[loss=0.2136, simple_loss=0.2936, pruned_loss=0.06677, over 1611292.13 frames. ], batch size: 22, lr: 4.16e-03, grad_scale: 8.0 +2023-02-06 22:18:53,261 INFO [optim.py:369] (0/4) 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 22:19:03,120 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.53 vs. limit=5.0 +2023-02-06 22:19:16,124 INFO [train.py:901] (0/4) Epoch 18, batch 7900, loss[loss=0.2045, simple_loss=0.2967, pruned_loss=0.05612, over 8448.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2935, pruned_loss=0.06611, over 1615205.20 frames. ], batch size: 24, lr: 4.16e-03, grad_scale: 8.0 +2023-02-06 22:19:28,936 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4709, 1.6956, 1.6833, 1.1049, 1.7446, 1.4082, 0.3002, 1.6127], + device='cuda:0'), covar=tensor([0.0367, 0.0298, 0.0260, 0.0394, 0.0333, 0.0755, 0.0732, 0.0216], + device='cuda:0'), in_proj_covar=tensor([0.0439, 0.0380, 0.0324, 0.0433, 0.0362, 0.0526, 0.0382, 0.0404], + device='cuda:0'), 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:0') +2023-02-06 22:19:44,153 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2391, 1.9546, 2.6615, 2.1724, 2.4704, 2.2704, 1.9675, 1.4617], + device='cuda:0'), covar=tensor([0.5174, 0.4657, 0.1693, 0.3353, 0.2459, 0.2937, 0.1966, 0.5007], + device='cuda:0'), in_proj_covar=tensor([0.0939, 0.0956, 0.0783, 0.0921, 0.0988, 0.0876, 0.0737, 0.0817], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 22:19:47,146 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 18, batch 7950, loss[loss=0.2493, simple_loss=0.3237, pruned_loss=0.08746, over 8236.00 frames. ], tot_loss[loss=0.2144, simple_loss=0.2949, pruned_loss=0.06693, over 1619172.00 frames. ], batch size: 24, lr: 4.16e-03, grad_scale: 8.0 +2023-02-06 22:19:58,003 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8445, 2.1336, 1.7483, 2.6038, 1.1722, 1.5413, 1.9256, 2.0302], + device='cuda:0'), covar=tensor([0.0735, 0.0673, 0.0875, 0.0380, 0.1167, 0.1269, 0.0829, 0.0727], + device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0197, 0.0248, 0.0210, 0.0206, 0.0244, 0.0250, 0.0209], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 22:19:59,835 INFO [optim.py:369] (0/4) 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 22:20:07,929 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=145389.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:20:08,001 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145389.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:20:23,107 INFO [train.py:901] (0/4) Epoch 18, batch 8000, loss[loss=0.24, simple_loss=0.3318, pruned_loss=0.07407, over 8102.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2931, pruned_loss=0.06562, over 1619936.66 frames. ], batch size: 23, lr: 4.16e-03, grad_scale: 8.0 +2023-02-06 22:20:34,646 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=145428.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:20:57,109 INFO [train.py:901] (0/4) Epoch 18, batch 8050, loss[loss=0.2261, simple_loss=0.2891, pruned_loss=0.08155, over 7536.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2919, pruned_loss=0.06603, over 1597166.35 frames. ], batch size: 18, lr: 4.16e-03, grad_scale: 8.0 +2023-02-06 22:21:05,673 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=145473.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:21:08,162 INFO [optim.py:369] (0/4) 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 22:21:19,730 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/epoch-18.pt +2023-02-06 22:21:30,964 WARNING [train.py:1067] (0/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-06 22:21:34,899 INFO [train.py:901] (0/4) Epoch 19, batch 0, loss[loss=0.2123, simple_loss=0.2933, pruned_loss=0.06571, over 8453.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2933, pruned_loss=0.06571, over 8453.00 frames. ], batch size: 29, lr: 4.05e-03, grad_scale: 8.0 +2023-02-06 22:21:34,900 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 22:21:46,554 INFO [train.py:935] (0/4) Epoch 19, validation: loss=0.1782, simple_loss=0.2779, pruned_loss=0.03928, over 944034.00 frames. +2023-02-06 22:21:46,555 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-06 22:21:54,190 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=145504.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:22:02,572 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.6465, 3.0694, 2.3683, 4.1026, 1.7815, 2.2786, 2.5750, 3.1681], + device='cuda:0'), covar=tensor([0.0614, 0.0731, 0.0896, 0.0236, 0.1060, 0.1105, 0.0894, 0.0740], + device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0199, 0.0250, 0.0212, 0.0208, 0.0246, 0.0252, 0.0210], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 22:22:03,065 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-06 22:22:22,466 INFO [train.py:901] (0/4) Epoch 19, batch 50, loss[loss=0.1647, simple_loss=0.243, pruned_loss=0.04318, over 7681.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2915, pruned_loss=0.06546, over 364260.31 frames. ], batch size: 18, lr: 4.05e-03, grad_scale: 8.0 +2023-02-06 22:22:22,665 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=145543.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:22:39,355 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.8598, 5.9925, 5.1621, 2.7752, 5.2416, 5.5814, 5.5393, 5.3006], + device='cuda:0'), covar=tensor([0.0479, 0.0359, 0.0894, 0.3959, 0.0738, 0.0649, 0.0924, 0.0626], + device='cuda:0'), in_proj_covar=tensor([0.0514, 0.0426, 0.0429, 0.0527, 0.0415, 0.0425, 0.0409, 0.0374], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 22:22:40,521 WARNING [train.py:1067] (0/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-06 22:22:45,198 INFO [optim.py:369] (0/4) 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 22:22:56,256 INFO [train.py:901] (0/4) Epoch 19, batch 100, loss[loss=0.2353, simple_loss=0.3054, pruned_loss=0.08256, over 7813.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.293, pruned_loss=0.06597, over 639021.87 frames. ], batch size: 20, lr: 4.05e-03, grad_scale: 8.0 +2023-02-06 22:23:01,896 WARNING [train.py:1067] (0/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-06 22:23:10,098 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145612.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:23:30,509 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.8320, 2.5257, 3.4838, 1.8160, 1.7229, 3.4731, 0.7557, 2.0566], + device='cuda:0'), covar=tensor([0.1583, 0.1069, 0.0230, 0.1876, 0.2992, 0.0341, 0.2195, 0.1472], + device='cuda:0'), in_proj_covar=tensor([0.0183, 0.0191, 0.0122, 0.0216, 0.0268, 0.0130, 0.0166, 0.0185], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 22:23:32,323 INFO [train.py:901] (0/4) Epoch 19, batch 150, loss[loss=0.1954, simple_loss=0.2732, pruned_loss=0.05883, over 7652.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.2929, pruned_loss=0.06604, over 857564.47 frames. ], batch size: 19, lr: 4.05e-03, grad_scale: 8.0 +2023-02-06 22:23:46,172 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145661.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:23:48,211 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.0213, 1.5812, 4.2223, 1.8251, 3.7805, 3.5221, 3.8661, 3.7225], + device='cuda:0'), covar=tensor([0.0751, 0.4664, 0.0621, 0.4104, 0.1128, 0.1032, 0.0649, 0.0728], + device='cuda:0'), in_proj_covar=tensor([0.0597, 0.0631, 0.0676, 0.0605, 0.0684, 0.0590, 0.0587, 0.0649], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 22:23:49,601 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1108, 2.3956, 1.8929, 2.8970, 1.4219, 1.7435, 2.1500, 2.4116], + device='cuda:0'), covar=tensor([0.0659, 0.0692, 0.0903, 0.0339, 0.1081, 0.1198, 0.0770, 0.0679], + device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0198, 0.0249, 0.0211, 0.0207, 0.0245, 0.0251, 0.0210], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 22:23:57,046 INFO [optim.py:369] (0/4) 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 22:24:07,965 INFO [train.py:901] (0/4) Epoch 19, batch 200, loss[loss=0.2152, simple_loss=0.3036, pruned_loss=0.06335, over 8574.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.2938, pruned_loss=0.06621, over 1028906.81 frames. ], batch size: 31, lr: 4.05e-03, grad_scale: 8.0 +2023-02-06 22:24:14,825 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9347, 1.6297, 3.2848, 1.4352, 2.2944, 3.5598, 3.7205, 3.0619], + device='cuda:0'), covar=tensor([0.1161, 0.1644, 0.0335, 0.2072, 0.1111, 0.0251, 0.0599, 0.0561], + device='cuda:0'), in_proj_covar=tensor([0.0284, 0.0314, 0.0277, 0.0305, 0.0295, 0.0255, 0.0399, 0.0296], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-06 22:24:33,110 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145729.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:24:35,721 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=145733.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:24:42,457 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 +2023-02-06 22:24:42,743 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9509, 3.8283, 2.3160, 2.8368, 2.8830, 2.0836, 3.0065, 2.9920], + device='cuda:0'), covar=tensor([0.1751, 0.0364, 0.1068, 0.0771, 0.0692, 0.1244, 0.1008, 0.1086], + device='cuda:0'), in_proj_covar=tensor([0.0351, 0.0234, 0.0323, 0.0300, 0.0296, 0.0328, 0.0339, 0.0313], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 22:24:43,207 INFO [train.py:901] (0/4) Epoch 19, batch 250, loss[loss=0.1983, simple_loss=0.2811, pruned_loss=0.05779, over 7922.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2943, pruned_loss=0.06623, over 1161271.62 frames. ], batch size: 20, lr: 4.04e-03, grad_scale: 8.0 +2023-02-06 22:24:51,126 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145754.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:24:55,229 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145760.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:24:58,388 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-06 22:25:06,968 WARNING [train.py:1067] (0/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-06 22:25:07,542 INFO [optim.py:369] (0/4) 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 22:25:07,762 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.1538, 2.1294, 1.5686, 1.9209, 1.7929, 1.2512, 1.6333, 1.5845], + device='cuda:0'), covar=tensor([0.1526, 0.0464, 0.1225, 0.0549, 0.0687, 0.1608, 0.1037, 0.0864], + device='cuda:0'), in_proj_covar=tensor([0.0352, 0.0235, 0.0323, 0.0301, 0.0296, 0.0329, 0.0340, 0.0313], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 22:25:13,288 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145785.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:25:17,453 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8232, 2.2906, 1.6912, 2.9790, 1.2837, 1.5179, 2.0543, 2.3939], + device='cuda:0'), covar=tensor([0.0934, 0.0811, 0.1193, 0.0372, 0.1218, 0.1484, 0.0859, 0.0681], + device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0197, 0.0248, 0.0211, 0.0207, 0.0245, 0.0251, 0.0209], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 22:25:18,657 INFO [train.py:901] (0/4) Epoch 19, batch 300, loss[loss=0.2018, simple_loss=0.2853, pruned_loss=0.05916, over 8081.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.294, pruned_loss=0.06635, over 1263629.27 frames. ], batch size: 21, lr: 4.04e-03, grad_scale: 8.0 +2023-02-06 22:25:20,818 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3736, 1.3513, 2.3730, 1.1921, 2.1694, 2.5012, 2.6597, 2.1278], + device='cuda:0'), covar=tensor([0.1139, 0.1375, 0.0443, 0.2096, 0.0777, 0.0385, 0.0614, 0.0675], + device='cuda:0'), in_proj_covar=tensor([0.0284, 0.0315, 0.0277, 0.0307, 0.0296, 0.0255, 0.0400, 0.0298], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-06 22:25:22,940 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145799.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:25:39,973 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145824.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:25:50,230 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145839.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 22:25:50,291 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7155, 1.9818, 2.2031, 1.4142, 2.2709, 1.6152, 0.6864, 1.9822], + device='cuda:0'), covar=tensor([0.0516, 0.0314, 0.0216, 0.0467, 0.0351, 0.0690, 0.0702, 0.0250], + device='cuda:0'), in_proj_covar=tensor([0.0437, 0.0378, 0.0324, 0.0431, 0.0361, 0.0522, 0.0380, 0.0404], + device='cuda:0'), out_proj_covar=tensor([1.1878e-04, 1.0002e-04, 8.5744e-05, 1.1446e-04, 9.6007e-05, 1.4906e-04, + 1.0307e-04, 1.0820e-04], device='cuda:0') +2023-02-06 22:25:53,699 INFO [train.py:901] (0/4) Epoch 19, batch 350, loss[loss=0.1774, simple_loss=0.2579, pruned_loss=0.04845, over 8181.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.2939, pruned_loss=0.06629, over 1345017.29 frames. ], batch size: 23, lr: 4.04e-03, grad_scale: 8.0 +2023-02-06 22:25:57,435 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=145848.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:26:16,559 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0620, 1.8465, 2.3446, 1.9975, 2.2852, 2.1523, 1.9097, 1.1699], + device='cuda:0'), covar=tensor([0.4912, 0.4220, 0.1622, 0.3382, 0.2143, 0.2481, 0.1747, 0.4592], + device='cuda:0'), in_proj_covar=tensor([0.0929, 0.0948, 0.0776, 0.0915, 0.0979, 0.0867, 0.0729, 0.0808], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 22:26:17,670 INFO [optim.py:369] (0/4) 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 22:26:30,034 INFO [train.py:901] (0/4) Epoch 19, batch 400, loss[loss=0.1791, simple_loss=0.2523, pruned_loss=0.05292, over 7781.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2915, pruned_loss=0.06516, over 1402902.07 frames. ], batch size: 19, lr: 4.04e-03, grad_scale: 8.0 +2023-02-06 22:27:04,033 INFO [train.py:901] (0/4) Epoch 19, batch 450, loss[loss=0.1999, simple_loss=0.2849, pruned_loss=0.05747, over 8181.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.2938, pruned_loss=0.06623, over 1453767.49 frames. ], batch size: 23, lr: 4.04e-03, grad_scale: 8.0 +2023-02-06 22:27:12,886 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=145956.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:27:28,524 INFO [optim.py:369] (0/4) 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 22:27:40,177 INFO [train.py:901] (0/4) Epoch 19, batch 500, loss[loss=0.188, simple_loss=0.2642, pruned_loss=0.05591, over 7638.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.294, pruned_loss=0.06656, over 1491743.47 frames. ], batch size: 19, lr: 4.04e-03, grad_scale: 8.0 +2023-02-06 22:27:45,786 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-146000.pt +2023-02-06 22:27:50,074 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=146005.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:28:03,398 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.6910, 1.5275, 4.8694, 1.7924, 4.3459, 3.9686, 4.4241, 4.2406], + device='cuda:0'), covar=tensor([0.0493, 0.4448, 0.0419, 0.3979, 0.0989, 0.0941, 0.0508, 0.0614], + device='cuda:0'), in_proj_covar=tensor([0.0597, 0.0627, 0.0673, 0.0605, 0.0687, 0.0591, 0.0585, 0.0649], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 22:28:15,839 INFO [train.py:901] (0/4) Epoch 19, batch 550, loss[loss=0.1746, simple_loss=0.2587, pruned_loss=0.0452, over 7938.00 frames. ], tot_loss[loss=0.2116, simple_loss=0.2922, pruned_loss=0.06545, over 1522548.79 frames. ], batch size: 20, lr: 4.04e-03, grad_scale: 8.0 +2023-02-06 22:28:19,280 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.5862, 5.7310, 5.0329, 2.3937, 5.0880, 5.4075, 5.3483, 5.0882], + device='cuda:0'), covar=tensor([0.0671, 0.0448, 0.1008, 0.4564, 0.0731, 0.0781, 0.1094, 0.0569], + device='cuda:0'), in_proj_covar=tensor([0.0519, 0.0428, 0.0430, 0.0530, 0.0416, 0.0430, 0.0411, 0.0375], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 22:28:35,097 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146071.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:28:38,933 INFO [optim.py:369] (0/4) 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 22:28:50,789 INFO [train.py:901] (0/4) Epoch 19, batch 600, loss[loss=0.2174, simple_loss=0.2971, pruned_loss=0.06884, over 8133.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2935, pruned_loss=0.06609, over 1546614.31 frames. ], batch size: 22, lr: 4.04e-03, grad_scale: 8.0 +2023-02-06 22:28:54,534 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.6560, 1.3618, 4.8723, 1.7339, 4.3295, 3.9976, 4.4256, 4.2439], + device='cuda:0'), covar=tensor([0.0499, 0.4887, 0.0466, 0.4046, 0.1063, 0.1015, 0.0520, 0.0635], + device='cuda:0'), in_proj_covar=tensor([0.0594, 0.0624, 0.0670, 0.0600, 0.0684, 0.0588, 0.0582, 0.0646], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 22:28:59,453 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=146104.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:29:03,461 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.4362, 4.3813, 4.0090, 2.1502, 3.9136, 4.0490, 4.0485, 3.7766], + device='cuda:0'), covar=tensor([0.0707, 0.0500, 0.1001, 0.4310, 0.0804, 0.1078, 0.1151, 0.0760], + device='cuda:0'), in_proj_covar=tensor([0.0517, 0.0425, 0.0427, 0.0526, 0.0413, 0.0427, 0.0407, 0.0373], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 22:29:11,428 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-06 22:29:11,605 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146120.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:29:17,640 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 19, batch 650, loss[loss=0.2061, simple_loss=0.2847, pruned_loss=0.06381, over 8034.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2925, pruned_loss=0.06582, over 1564083.19 frames. ], batch size: 22, lr: 4.04e-03, grad_scale: 8.0 +2023-02-06 22:29:42,897 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=146167.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:29:49,756 INFO [optim.py:369] (0/4) 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 22:29:53,905 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=146183.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 22:30:00,618 INFO [train.py:901] (0/4) Epoch 19, batch 700, loss[loss=0.2125, simple_loss=0.2921, pruned_loss=0.06648, over 8080.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2931, pruned_loss=0.06635, over 1574790.12 frames. ], batch size: 21, lr: 4.04e-03, grad_scale: 8.0 +2023-02-06 22:30:37,716 INFO [train.py:901] (0/4) Epoch 19, batch 750, loss[loss=0.2103, simple_loss=0.2926, pruned_loss=0.06398, over 8023.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2917, pruned_loss=0.06518, over 1582685.47 frames. ], batch size: 22, lr: 4.04e-03, grad_scale: 8.0 +2023-02-06 22:30:58,054 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-06 22:31:00,738 INFO [optim.py:369] (0/4) 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 22:31:06,862 WARNING [train.py:1067] (0/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-06 22:31:11,496 INFO [train.py:901] (0/4) Epoch 19, batch 800, loss[loss=0.1893, simple_loss=0.2695, pruned_loss=0.05455, over 7806.00 frames. ], tot_loss[loss=0.2126, simple_loss=0.293, pruned_loss=0.06613, over 1591803.32 frames. ], batch size: 20, lr: 4.04e-03, grad_scale: 8.0 +2023-02-06 22:31:14,849 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146298.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 22:31:15,523 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1944, 1.6028, 1.7886, 1.4418, 1.0003, 1.5828, 1.9043, 1.7971], + device='cuda:0'), covar=tensor([0.0455, 0.1193, 0.1593, 0.1365, 0.0572, 0.1423, 0.0630, 0.0597], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0153, 0.0190, 0.0158, 0.0099, 0.0162, 0.0113, 0.0141], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 22:31:17,173 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.53 vs. limit=5.0 +2023-02-06 22:31:35,761 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 19, batch 850, loss[loss=0.1787, simple_loss=0.2637, pruned_loss=0.04689, over 8135.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2922, pruned_loss=0.06577, over 1592393.88 frames. ], batch size: 22, lr: 4.04e-03, grad_scale: 8.0 +2023-02-06 22:31:53,497 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=146352.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:32:10,849 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=146376.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:32:11,293 INFO [optim.py:369] (0/4) 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 22:32:22,245 INFO [train.py:901] (0/4) Epoch 19, batch 900, loss[loss=0.2421, simple_loss=0.318, pruned_loss=0.0831, over 8749.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2929, pruned_loss=0.06568, over 1595214.44 frames. ], batch size: 49, lr: 4.04e-03, grad_scale: 8.0 +2023-02-06 22:32:27,825 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=146401.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:32:41,206 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6526, 1.9309, 3.1554, 1.4115, 2.4077, 2.0698, 1.7391, 2.3448], + device='cuda:0'), covar=tensor([0.1880, 0.2719, 0.0902, 0.4735, 0.1702, 0.3117, 0.2276, 0.2145], + device='cuda:0'), in_proj_covar=tensor([0.0516, 0.0580, 0.0550, 0.0630, 0.0639, 0.0584, 0.0518, 0.0629], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 22:32:56,371 INFO [train.py:901] (0/4) Epoch 19, batch 950, loss[loss=0.229, simple_loss=0.3107, pruned_loss=0.07367, over 8492.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.2936, pruned_loss=0.06599, over 1601942.19 frames. ], batch size: 29, lr: 4.04e-03, grad_scale: 8.0 +2023-02-06 22:33:09,722 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.46 vs. limit=5.0 +2023-02-06 22:33:20,834 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1529, 1.5075, 1.7234, 1.3613, 0.9320, 1.4096, 1.6575, 1.4557], + device='cuda:0'), covar=tensor([0.0491, 0.1217, 0.1617, 0.1412, 0.0613, 0.1542, 0.0703, 0.0697], + device='cuda:0'), in_proj_covar=tensor([0.0096, 0.0151, 0.0188, 0.0157, 0.0099, 0.0162, 0.0112, 0.0140], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 22:33:21,319 INFO [optim.py:369] (0/4) 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 22:33:22,702 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-06 22:33:32,161 INFO [train.py:901] (0/4) Epoch 19, batch 1000, loss[loss=0.1755, simple_loss=0.2581, pruned_loss=0.0465, over 7554.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2942, pruned_loss=0.06639, over 1611586.02 frames. ], batch size: 18, lr: 4.03e-03, grad_scale: 8.0 +2023-02-06 22:33:44,471 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=146511.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:33:54,580 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-06 22:34:06,367 INFO [train.py:901] (0/4) Epoch 19, batch 1050, loss[loss=0.2288, simple_loss=0.3126, pruned_loss=0.07247, over 8513.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2937, pruned_loss=0.06585, over 1612773.03 frames. ], batch size: 28, lr: 4.03e-03, grad_scale: 8.0 +2023-02-06 22:34:06,384 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-06 22:34:14,932 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=146554.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 22:34:31,581 INFO [optim.py:369] (0/4) 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 22:34:33,864 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 19, batch 1100, loss[loss=0.2196, simple_loss=0.2861, pruned_loss=0.07658, over 7931.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2926, pruned_loss=0.06591, over 1606874.09 frames. ], batch size: 20, lr: 4.03e-03, grad_scale: 8.0 +2023-02-06 22:34:55,179 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=146609.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:34:56,640 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2424, 2.0528, 2.8786, 2.2983, 2.6601, 2.2820, 1.9795, 1.4752], + device='cuda:0'), covar=tensor([0.4867, 0.4616, 0.1706, 0.3385, 0.2354, 0.2825, 0.1814, 0.5229], + device='cuda:0'), in_proj_covar=tensor([0.0931, 0.0948, 0.0782, 0.0913, 0.0978, 0.0866, 0.0726, 0.0809], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 22:35:06,983 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 19, batch 1150, loss[loss=0.1969, simple_loss=0.2767, pruned_loss=0.0585, over 7826.00 frames. ], tot_loss[loss=0.2114, simple_loss=0.292, pruned_loss=0.06534, over 1608352.61 frames. ], batch size: 20, lr: 4.03e-03, grad_scale: 8.0 +2023-02-06 22:35:19,108 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-06 22:35:19,253 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=146644.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:35:42,421 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 19, batch 1200, loss[loss=0.2172, simple_loss=0.2987, pruned_loss=0.06781, over 8453.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2927, pruned_loss=0.06546, over 1614610.18 frames. ], batch size: 25, lr: 4.03e-03, grad_scale: 8.0 +2023-02-06 22:36:11,193 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8285, 1.8174, 2.4309, 1.6750, 1.3052, 2.4802, 0.4223, 1.4946], + device='cuda:0'), covar=tensor([0.2048, 0.1250, 0.0348, 0.1401, 0.2857, 0.0372, 0.2272, 0.1498], + device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0192, 0.0123, 0.0218, 0.0267, 0.0130, 0.0168, 0.0185], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 22:36:29,002 INFO [train.py:901] (0/4) Epoch 19, batch 1250, loss[loss=0.214, simple_loss=0.2782, pruned_loss=0.07492, over 7691.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.2928, pruned_loss=0.06606, over 1614016.94 frames. ], batch size: 18, lr: 4.03e-03, grad_scale: 8.0 +2023-02-06 22:36:31,924 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2402, 2.6561, 2.8009, 1.9369, 3.1589, 1.9630, 1.5241, 2.1843], + device='cuda:0'), covar=tensor([0.0816, 0.0376, 0.0290, 0.0673, 0.0392, 0.0786, 0.0878, 0.0515], + device='cuda:0'), in_proj_covar=tensor([0.0435, 0.0378, 0.0324, 0.0429, 0.0361, 0.0521, 0.0381, 0.0403], + device='cuda:0'), out_proj_covar=tensor([1.1803e-04, 9.9890e-05, 8.5681e-05, 1.1387e-04, 9.6013e-05, 1.4899e-04, + 1.0310e-04, 1.0761e-04], device='cuda:0') +2023-02-06 22:36:52,652 INFO [optim.py:369] (0/4) 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 22:37:04,310 INFO [train.py:901] (0/4) Epoch 19, batch 1300, loss[loss=0.2271, simple_loss=0.2941, pruned_loss=0.0801, over 8071.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2931, pruned_loss=0.06612, over 1616614.62 frames. ], batch size: 21, lr: 4.03e-03, grad_scale: 8.0 +2023-02-06 22:37:40,717 INFO [train.py:901] (0/4) Epoch 19, batch 1350, loss[loss=0.1887, simple_loss=0.259, pruned_loss=0.05916, over 7712.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2923, pruned_loss=0.06556, over 1619714.19 frames. ], batch size: 18, lr: 4.03e-03, grad_scale: 8.0 +2023-02-06 22:37:53,728 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=146862.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:38:03,896 INFO [optim.py:369] (0/4) 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 22:38:07,780 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=146882.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:38:14,731 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.4276, 4.4729, 4.0057, 2.2319, 3.9436, 4.0052, 4.0509, 3.8075], + device='cuda:0'), covar=tensor([0.0702, 0.0558, 0.1104, 0.3875, 0.0834, 0.0857, 0.1231, 0.0853], + device='cuda:0'), in_proj_covar=tensor([0.0519, 0.0426, 0.0432, 0.0528, 0.0417, 0.0429, 0.0412, 0.0378], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 22:38:15,248 INFO [train.py:901] (0/4) Epoch 19, batch 1400, loss[loss=0.2243, simple_loss=0.3016, pruned_loss=0.07349, over 8126.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2913, pruned_loss=0.06517, over 1615787.02 frames. ], batch size: 22, lr: 4.03e-03, grad_scale: 8.0 +2023-02-06 22:38:19,630 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2444, 1.4761, 1.4951, 1.0247, 1.4749, 1.1551, 0.3290, 1.3954], + device='cuda:0'), covar=tensor([0.0502, 0.0397, 0.0305, 0.0488, 0.0470, 0.0866, 0.0822, 0.0273], + device='cuda:0'), in_proj_covar=tensor([0.0435, 0.0377, 0.0323, 0.0429, 0.0360, 0.0521, 0.0380, 0.0402], + device='cuda:0'), out_proj_covar=tensor([1.1802e-04, 9.9697e-05, 8.5509e-05, 1.1409e-04, 9.5812e-05, 1.4897e-04, + 1.0291e-04, 1.0748e-04], device='cuda:0') +2023-02-06 22:38:25,965 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=146907.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:38:43,247 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 +2023-02-06 22:38:52,633 INFO [train.py:901] (0/4) Epoch 19, batch 1450, loss[loss=0.2136, simple_loss=0.3041, pruned_loss=0.06151, over 8326.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2909, pruned_loss=0.06431, over 1619738.83 frames. ], batch size: 25, lr: 4.03e-03, grad_scale: 16.0 +2023-02-06 22:38:56,650 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-06 22:38:59,393 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=146953.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:39:16,190 INFO [optim.py:369] (0/4) 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 22:39:22,617 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.6773, 1.2734, 4.8343, 1.7963, 4.3022, 4.0207, 4.3903, 4.2586], + device='cuda:0'), covar=tensor([0.0555, 0.5014, 0.0464, 0.4113, 0.1081, 0.1016, 0.0567, 0.0647], + device='cuda:0'), in_proj_covar=tensor([0.0602, 0.0635, 0.0677, 0.0609, 0.0692, 0.0594, 0.0590, 0.0651], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 22:39:23,965 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=146988.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:39:27,299 INFO [train.py:901] (0/4) Epoch 19, batch 1500, loss[loss=0.2026, simple_loss=0.2847, pruned_loss=0.06026, over 8353.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2915, pruned_loss=0.06458, over 1620423.70 frames. ], batch size: 24, lr: 4.03e-03, grad_scale: 16.0 +2023-02-06 22:40:03,264 INFO [train.py:901] (0/4) Epoch 19, batch 1550, loss[loss=0.2228, simple_loss=0.2996, pruned_loss=0.07298, over 8136.00 frames. ], tot_loss[loss=0.2107, simple_loss=0.2921, pruned_loss=0.06463, over 1621780.02 frames. ], batch size: 22, lr: 4.03e-03, grad_scale: 16.0 +2023-02-06 22:40:22,631 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147068.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:40:28,393 INFO [optim.py:369] (0/4) 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 22:40:39,453 INFO [train.py:901] (0/4) Epoch 19, batch 1600, loss[loss=0.1754, simple_loss=0.2592, pruned_loss=0.04583, over 7203.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2913, pruned_loss=0.06397, over 1621113.88 frames. ], batch size: 16, lr: 4.03e-03, grad_scale: 16.0 +2023-02-06 22:40:46,376 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 19, batch 1650, loss[loss=0.2362, simple_loss=0.3079, pruned_loss=0.08222, over 8307.00 frames. ], tot_loss[loss=0.211, simple_loss=0.292, pruned_loss=0.06503, over 1621087.59 frames. ], batch size: 48, lr: 4.03e-03, grad_scale: 8.0 +2023-02-06 22:41:40,969 INFO [optim.py:369] (0/4) 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 22:41:51,231 INFO [train.py:901] (0/4) Epoch 19, batch 1700, loss[loss=0.2212, simple_loss=0.3029, pruned_loss=0.06977, over 8367.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2913, pruned_loss=0.06441, over 1619179.00 frames. ], batch size: 49, lr: 4.02e-03, grad_scale: 8.0 +2023-02-06 22:41:52,204 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1165, 1.8201, 2.3213, 1.9907, 2.2715, 2.1309, 1.8938, 1.1744], + device='cuda:0'), covar=tensor([0.4955, 0.4281, 0.1753, 0.3053, 0.2111, 0.2638, 0.1761, 0.4523], + device='cuda:0'), in_proj_covar=tensor([0.0936, 0.0953, 0.0788, 0.0917, 0.0984, 0.0869, 0.0730, 0.0810], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 22:42:00,560 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=147206.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:42:25,937 INFO [train.py:901] (0/4) Epoch 19, batch 1750, loss[loss=0.2171, simple_loss=0.3109, pruned_loss=0.06162, over 8241.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2911, pruned_loss=0.06424, over 1615561.43 frames. ], batch size: 24, lr: 4.02e-03, grad_scale: 8.0 +2023-02-06 22:42:37,947 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147259.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:42:41,391 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.5355, 1.3658, 4.7106, 1.8737, 4.1887, 3.9188, 4.2619, 4.1433], + device='cuda:0'), covar=tensor([0.0528, 0.4812, 0.0488, 0.4006, 0.0970, 0.0972, 0.0530, 0.0604], + device='cuda:0'), in_proj_covar=tensor([0.0601, 0.0636, 0.0677, 0.0612, 0.0692, 0.0596, 0.0591, 0.0653], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 22:42:46,351 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.6881, 4.7304, 4.1709, 2.0552, 4.1566, 4.2949, 4.2687, 4.1481], + device='cuda:0'), covar=tensor([0.0662, 0.0555, 0.1146, 0.4426, 0.0815, 0.0979, 0.1337, 0.0839], + device='cuda:0'), in_proj_covar=tensor([0.0513, 0.0423, 0.0428, 0.0526, 0.0415, 0.0427, 0.0410, 0.0375], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 22:42:51,045 INFO [optim.py:369] (0/4) 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 22:43:03,029 INFO [train.py:901] (0/4) Epoch 19, batch 1800, loss[loss=0.2106, simple_loss=0.2983, pruned_loss=0.06145, over 8238.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2912, pruned_loss=0.06451, over 1615832.00 frames. ], batch size: 22, lr: 4.02e-03, grad_scale: 8.0 +2023-02-06 22:43:22,531 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147321.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:43:24,589 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147324.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:43:27,334 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147328.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:43:27,658 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 +2023-02-06 22:43:37,468 INFO [train.py:901] (0/4) Epoch 19, batch 1850, loss[loss=0.2421, simple_loss=0.3259, pruned_loss=0.07918, over 8293.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.292, pruned_loss=0.06494, over 1616583.75 frames. ], batch size: 23, lr: 4.02e-03, grad_scale: 8.0 +2023-02-06 22:43:41,689 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147349.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:43:48,528 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147359.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:44:02,398 INFO [optim.py:369] (0/4) 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] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-02-06 22:44:06,632 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147384.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:44:12,522 INFO [train.py:901] (0/4) Epoch 19, batch 1900, loss[loss=0.2085, simple_loss=0.2861, pruned_loss=0.06547, over 8246.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2919, pruned_loss=0.06486, over 1614291.36 frames. ], batch size: 22, lr: 4.02e-03, grad_scale: 8.0 +2023-02-06 22:44:37,268 INFO [zipformer.py:1185] (0/4) 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] (0/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-06 22:44:49,675 INFO [train.py:901] (0/4) Epoch 19, batch 1950, loss[loss=0.2212, simple_loss=0.2944, pruned_loss=0.07401, over 7675.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2906, pruned_loss=0.06388, over 1615255.07 frames. ], batch size: 18, lr: 4.02e-03, grad_scale: 8.0 +2023-02-06 22:44:55,957 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147452.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:44:56,496 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-06 22:45:13,741 INFO [optim.py:369] (0/4) 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 22:45:15,252 WARNING [train.py:1067] (0/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-06 22:45:24,858 INFO [train.py:901] (0/4) Epoch 19, batch 2000, loss[loss=0.2389, simple_loss=0.3188, pruned_loss=0.07953, over 8550.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2915, pruned_loss=0.06468, over 1611300.38 frames. ], batch size: 49, lr: 4.02e-03, grad_scale: 8.0 +2023-02-06 22:46:01,778 INFO [train.py:901] (0/4) Epoch 19, batch 2050, loss[loss=0.2076, simple_loss=0.2817, pruned_loss=0.06677, over 7929.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2917, pruned_loss=0.06471, over 1613942.15 frames. ], batch size: 20, lr: 4.02e-03, grad_scale: 8.0 +2023-02-06 22:46:09,084 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.01 vs. limit=5.0 +2023-02-06 22:46:25,327 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 22:46:36,254 INFO [train.py:901] (0/4) Epoch 19, batch 2100, loss[loss=0.1816, simple_loss=0.2663, pruned_loss=0.04844, over 7979.00 frames. ], tot_loss[loss=0.211, simple_loss=0.292, pruned_loss=0.06499, over 1612653.92 frames. ], batch size: 21, lr: 4.02e-03, grad_scale: 8.0 +2023-02-06 22:46:42,937 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147602.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:46:43,425 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=147603.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:46:47,319 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.65 vs. limit=5.0 +2023-02-06 22:47:12,100 INFO [train.py:901] (0/4) Epoch 19, batch 2150, loss[loss=0.2119, simple_loss=0.298, pruned_loss=0.06284, over 8138.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2915, pruned_loss=0.06501, over 1611727.43 frames. ], batch size: 22, lr: 4.02e-03, grad_scale: 8.0 +2023-02-06 22:47:28,862 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2780, 2.1949, 1.7139, 1.9160, 1.7346, 1.5190, 1.7262, 1.6495], + device='cuda:0'), covar=tensor([0.1239, 0.0387, 0.1088, 0.0563, 0.0757, 0.1354, 0.0835, 0.0794], + device='cuda:0'), in_proj_covar=tensor([0.0353, 0.0234, 0.0325, 0.0303, 0.0301, 0.0331, 0.0341, 0.0316], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 22:47:31,898 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 +2023-02-06 22:47:33,534 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=147672.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:47:36,909 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 22:47:47,723 INFO [train.py:901] (0/4) Epoch 19, batch 2200, loss[loss=0.1719, simple_loss=0.2522, pruned_loss=0.04577, over 7651.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2915, pruned_loss=0.06544, over 1604663.53 frames. ], batch size: 19, lr: 4.02e-03, grad_scale: 8.0 +2023-02-06 22:48:04,723 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147718.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:48:21,942 INFO [train.py:901] (0/4) Epoch 19, batch 2250, loss[loss=0.21, simple_loss=0.302, pruned_loss=0.05898, over 8466.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.293, pruned_loss=0.06603, over 1607259.16 frames. ], batch size: 25, lr: 4.02e-03, grad_scale: 8.0 +2023-02-06 22:48:41,105 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 22:48:53,325 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 19, batch 2300, loss[loss=0.1638, simple_loss=0.2493, pruned_loss=0.03911, over 8079.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.2931, pruned_loss=0.06623, over 1608225.23 frames. ], batch size: 21, lr: 4.02e-03, grad_scale: 8.0 +2023-02-06 22:48:58,973 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=147796.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:49:24,328 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147833.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:49:30,985 INFO [train.py:901] (0/4) Epoch 19, batch 2350, loss[loss=0.2194, simple_loss=0.3035, pruned_loss=0.0676, over 8595.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.2939, pruned_loss=0.06631, over 1613441.10 frames. ], batch size: 31, lr: 4.02e-03, grad_scale: 8.0 +2023-02-06 22:49:53,245 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147875.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:49:55,888 INFO [optim.py:369] (0/4) 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 22:50:01,527 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147884.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:50:04,972 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.8339, 2.9731, 2.5988, 4.1362, 1.8123, 2.3145, 2.5890, 3.3971], + device='cuda:0'), covar=tensor([0.0562, 0.0761, 0.0752, 0.0176, 0.1073, 0.1089, 0.0994, 0.0592], + device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0197, 0.0249, 0.0212, 0.0207, 0.0246, 0.0253, 0.0212], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 22:50:07,513 INFO [train.py:901] (0/4) Epoch 19, batch 2400, loss[loss=0.195, simple_loss=0.2648, pruned_loss=0.06255, over 7435.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2921, pruned_loss=0.06521, over 1612955.26 frames. ], batch size: 17, lr: 4.02e-03, grad_scale: 8.0 +2023-02-06 22:50:17,626 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.68 vs. limit=5.0 +2023-02-06 22:50:20,055 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147911.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:50:32,258 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147929.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:50:41,459 INFO [train.py:901] (0/4) Epoch 19, batch 2450, loss[loss=0.1845, simple_loss=0.2709, pruned_loss=0.04902, over 8108.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2922, pruned_loss=0.06561, over 1612454.30 frames. ], batch size: 23, lr: 4.01e-03, grad_scale: 8.0 +2023-02-06 22:51:03,062 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 22:51:15,596 INFO [train.py:901] (0/4) Epoch 19, batch 2500, loss[loss=0.2027, simple_loss=0.285, pruned_loss=0.06027, over 8333.00 frames. ], tot_loss[loss=0.2114, simple_loss=0.2922, pruned_loss=0.0653, over 1615930.79 frames. ], batch size: 26, lr: 4.01e-03, grad_scale: 8.0 +2023-02-06 22:51:20,625 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147999.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:51:21,200 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-148000.pt +2023-02-06 22:51:37,735 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=148021.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:51:41,216 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8104, 2.4809, 4.0925, 1.6836, 3.1263, 2.2780, 2.0796, 2.7134], + device='cuda:0'), covar=tensor([0.1759, 0.2229, 0.0821, 0.4054, 0.1542, 0.3040, 0.1850, 0.2459], + device='cuda:0'), in_proj_covar=tensor([0.0513, 0.0577, 0.0548, 0.0626, 0.0634, 0.0583, 0.0518, 0.0627], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 22:51:52,333 INFO [train.py:901] (0/4) Epoch 19, batch 2550, loss[loss=0.2145, simple_loss=0.3019, pruned_loss=0.06354, over 8629.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2917, pruned_loss=0.06522, over 1615620.57 frames. ], batch size: 49, lr: 4.01e-03, grad_scale: 8.0 +2023-02-06 22:51:52,608 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148043.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:52:09,266 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148068.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:52:15,629 INFO [optim.py:369] (0/4) 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 22:52:17,123 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2979, 2.7186, 3.2018, 1.7423, 3.1491, 1.9090, 1.7168, 2.3847], + device='cuda:0'), covar=tensor([0.0821, 0.0391, 0.0290, 0.0832, 0.0502, 0.0957, 0.0833, 0.0419], + device='cuda:0'), in_proj_covar=tensor([0.0441, 0.0380, 0.0329, 0.0435, 0.0363, 0.0526, 0.0383, 0.0404], + device='cuda:0'), 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:0') +2023-02-06 22:52:24,502 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2933, 2.3335, 1.6918, 2.0486, 1.7939, 1.3670, 1.7677, 1.8402], + device='cuda:0'), covar=tensor([0.1445, 0.0419, 0.1253, 0.0588, 0.0842, 0.1625, 0.1075, 0.0941], + device='cuda:0'), in_proj_covar=tensor([0.0354, 0.0236, 0.0326, 0.0305, 0.0302, 0.0333, 0.0343, 0.0318], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 22:52:26,269 INFO [train.py:901] (0/4) Epoch 19, batch 2600, loss[loss=0.2239, simple_loss=0.3106, pruned_loss=0.06862, over 8238.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2931, pruned_loss=0.06676, over 1612537.26 frames. ], batch size: 24, lr: 4.01e-03, grad_scale: 8.0 +2023-02-06 22:52:57,239 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148140.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:53:01,844 INFO [train.py:901] (0/4) Epoch 19, batch 2650, loss[loss=0.22, simple_loss=0.3104, pruned_loss=0.0648, over 8368.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.2942, pruned_loss=0.06745, over 1614067.55 frames. ], batch size: 24, lr: 4.01e-03, grad_scale: 8.0 +2023-02-06 22:53:16,796 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148165.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:53:18,187 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148167.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:53:24,871 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.47 vs. limit=2.0 +2023-02-06 22:53:35,175 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 19, batch 2700, loss[loss=0.212, simple_loss=0.2988, pruned_loss=0.06258, over 8537.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.2933, pruned_loss=0.0665, over 1614561.42 frames. ], batch size: 34, lr: 4.01e-03, grad_scale: 8.0 +2023-02-06 22:53:38,566 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5913, 2.6939, 1.9106, 2.4393, 2.2786, 1.7142, 2.2490, 2.2496], + device='cuda:0'), covar=tensor([0.1577, 0.0404, 0.1116, 0.0615, 0.0838, 0.1387, 0.0965, 0.1009], + device='cuda:0'), in_proj_covar=tensor([0.0354, 0.0237, 0.0327, 0.0306, 0.0302, 0.0333, 0.0344, 0.0318], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 22:53:54,165 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=148219.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:54:03,916 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8133, 2.0028, 1.7059, 2.3585, 1.1035, 1.5501, 1.7461, 1.9318], + device='cuda:0'), covar=tensor([0.0793, 0.0650, 0.1022, 0.0486, 0.1025, 0.1309, 0.0769, 0.0663], + device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0198, 0.0252, 0.0214, 0.0208, 0.0248, 0.0256, 0.0213], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 22:54:11,923 INFO [train.py:901] (0/4) Epoch 19, batch 2750, loss[loss=0.1919, simple_loss=0.2662, pruned_loss=0.05875, over 7411.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2937, pruned_loss=0.06654, over 1615229.47 frames. ], batch size: 17, lr: 4.01e-03, grad_scale: 8.0 +2023-02-06 22:54:32,992 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=148273.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:54:36,059 INFO [optim.py:369] (0/4) 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] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148292.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:54:45,919 INFO [train.py:901] (0/4) Epoch 19, batch 2800, loss[loss=0.2298, simple_loss=0.3071, pruned_loss=0.0762, over 8086.00 frames. ], tot_loss[loss=0.2136, simple_loss=0.2938, pruned_loss=0.06672, over 1613356.33 frames. ], batch size: 21, lr: 4.01e-03, grad_scale: 8.0 +2023-02-06 22:54:54,036 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148305.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:55:13,964 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148334.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:55:19,842 INFO [train.py:901] (0/4) Epoch 19, batch 2850, loss[loss=0.195, simple_loss=0.277, pruned_loss=0.05648, over 7797.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.2931, pruned_loss=0.06623, over 1613049.26 frames. ], batch size: 20, lr: 4.01e-03, grad_scale: 8.0 +2023-02-06 22:55:29,871 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.85 vs. limit=5.0 +2023-02-06 22:55:46,065 INFO [optim.py:369] (0/4) 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 22:55:52,982 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148388.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:55:55,623 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 19, batch 2900, loss[loss=0.2294, simple_loss=0.3229, pruned_loss=0.06789, over 8506.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2937, pruned_loss=0.06609, over 1617086.91 frames. ], batch size: 28, lr: 4.01e-03, grad_scale: 8.0 +2023-02-06 22:56:10,223 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 +2023-02-06 22:56:12,655 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148417.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:56:29,355 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-06 22:56:29,948 INFO [train.py:901] (0/4) Epoch 19, batch 2950, loss[loss=0.2037, simple_loss=0.2765, pruned_loss=0.06546, over 8085.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2939, pruned_loss=0.06645, over 1615496.09 frames. ], batch size: 21, lr: 4.01e-03, grad_scale: 8.0 +2023-02-06 22:56:32,795 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148447.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:56:52,406 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1085, 1.6522, 1.7000, 1.4703, 1.0175, 1.5351, 1.8115, 1.5940], + device='cuda:0'), covar=tensor([0.0500, 0.1183, 0.1616, 0.1405, 0.0598, 0.1475, 0.0681, 0.0615], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0153, 0.0190, 0.0158, 0.0100, 0.0162, 0.0113, 0.0141], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 22:56:54,932 INFO [optim.py:369] (0/4) 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 22:57:06,356 INFO [train.py:901] (0/4) Epoch 19, batch 3000, loss[loss=0.2858, simple_loss=0.3543, pruned_loss=0.1086, over 6665.00 frames. ], tot_loss[loss=0.2137, simple_loss=0.2938, pruned_loss=0.06674, over 1613327.87 frames. ], batch size: 72, lr: 4.01e-03, grad_scale: 8.0 +2023-02-06 22:57:06,357 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 22:57:22,669 INFO [train.py:935] (0/4) Epoch 19, validation: loss=0.1752, simple_loss=0.2756, pruned_loss=0.03738, over 944034.00 frames. +2023-02-06 22:57:22,671 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-06 22:57:26,393 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9378, 1.8280, 2.9334, 2.2611, 2.5594, 1.9208, 1.6327, 1.2712], + device='cuda:0'), covar=tensor([0.6632, 0.5598, 0.1681, 0.3842, 0.2918, 0.4179, 0.2945, 0.5310], + device='cuda:0'), in_proj_covar=tensor([0.0932, 0.0952, 0.0782, 0.0916, 0.0980, 0.0868, 0.0730, 0.0809], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 22:57:38,582 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148516.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:57:44,374 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-02-06 22:57:56,924 INFO [train.py:901] (0/4) Epoch 19, batch 3050, loss[loss=0.2561, simple_loss=0.3244, pruned_loss=0.09389, over 8678.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.2933, pruned_loss=0.06648, over 1615131.96 frames. ], batch size: 39, lr: 4.01e-03, grad_scale: 8.0 +2023-02-06 22:58:00,725 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148573.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:58:21,053 INFO [optim.py:369] (0/4) 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 22:58:30,231 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148590.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:58:31,671 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3212, 2.6217, 3.0820, 1.7508, 3.3146, 2.1444, 1.5508, 2.2036], + device='cuda:0'), covar=tensor([0.0812, 0.0443, 0.0272, 0.0821, 0.0373, 0.0783, 0.0888, 0.0614], + device='cuda:0'), in_proj_covar=tensor([0.0442, 0.0380, 0.0330, 0.0437, 0.0364, 0.0531, 0.0384, 0.0406], + device='cuda:0'), out_proj_covar=tensor([1.1984e-04, 1.0038e-04, 8.7138e-05, 1.1622e-04, 9.6670e-05, 1.5179e-04, + 1.0388e-04, 1.0840e-04], device='cuda:0') +2023-02-06 22:58:32,155 INFO [train.py:901] (0/4) Epoch 19, batch 3100, loss[loss=0.1797, simple_loss=0.2644, pruned_loss=0.04753, over 8096.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.2932, pruned_loss=0.06649, over 1614994.02 frames. ], batch size: 21, lr: 4.01e-03, grad_scale: 8.0 +2023-02-06 22:58:33,712 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.0112, 2.6215, 3.6680, 2.0628, 2.1334, 3.6026, 0.6585, 2.2527], + device='cuda:0'), covar=tensor([0.1685, 0.1428, 0.0226, 0.1905, 0.2557, 0.0346, 0.2757, 0.1559], + device='cuda:0'), in_proj_covar=tensor([0.0183, 0.0192, 0.0122, 0.0218, 0.0267, 0.0130, 0.0168, 0.0185], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 22:58:49,317 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148615.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:58:50,729 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.5197, 2.9821, 2.4856, 3.9865, 1.8292, 2.2097, 2.5034, 3.1895], + device='cuda:0'), covar=tensor([0.0721, 0.0814, 0.0756, 0.0196, 0.1133, 0.1213, 0.1039, 0.0714], + device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0197, 0.0249, 0.0212, 0.0206, 0.0246, 0.0253, 0.0210], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 22:59:09,444 INFO [train.py:901] (0/4) Epoch 19, batch 3150, loss[loss=0.2305, simple_loss=0.3232, pruned_loss=0.06887, over 8133.00 frames. ], tot_loss[loss=0.212, simple_loss=0.293, pruned_loss=0.06546, over 1614783.57 frames. ], batch size: 22, lr: 4.01e-03, grad_scale: 8.0 +2023-02-06 22:59:10,311 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148644.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:59:13,423 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=148649.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:59:26,331 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148669.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 22:59:26,394 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 22:59:42,395 INFO [train.py:901] (0/4) Epoch 19, batch 3200, loss[loss=0.2665, simple_loss=0.3384, pruned_loss=0.09729, over 8473.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.293, pruned_loss=0.0656, over 1615613.10 frames. ], batch size: 25, lr: 4.00e-03, grad_scale: 8.0 +2023-02-06 23:00:12,966 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 19, batch 3250, loss[loss=0.1889, simple_loss=0.2679, pruned_loss=0.0549, over 7820.00 frames. ], tot_loss[loss=0.2126, simple_loss=0.2932, pruned_loss=0.06601, over 1613826.25 frames. ], batch size: 20, lr: 4.00e-03, grad_scale: 8.0 +2023-02-06 23:00:26,676 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0574, 1.4317, 1.6122, 1.3950, 0.8775, 1.4498, 1.7025, 1.5994], + device='cuda:0'), covar=tensor([0.0517, 0.1341, 0.1801, 0.1490, 0.0643, 0.1530, 0.0738, 0.0629], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0154, 0.0191, 0.0158, 0.0100, 0.0162, 0.0113, 0.0141], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 23:00:34,076 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148764.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:00:43,259 INFO [optim.py:369] (0/4) 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 23:00:52,324 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=148791.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:00:53,603 INFO [train.py:901] (0/4) Epoch 19, batch 3300, loss[loss=0.2175, simple_loss=0.3093, pruned_loss=0.06282, over 7808.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2914, pruned_loss=0.06515, over 1610291.61 frames. ], batch size: 20, lr: 4.00e-03, grad_scale: 8.0 +2023-02-06 23:01:28,250 INFO [train.py:901] (0/4) Epoch 19, batch 3350, loss[loss=0.2184, simple_loss=0.2965, pruned_loss=0.0701, over 8363.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2927, pruned_loss=0.06599, over 1608403.76 frames. ], batch size: 24, lr: 4.00e-03, grad_scale: 8.0 +2023-02-06 23:01:31,900 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4868, 2.2714, 3.1402, 2.4473, 3.0099, 2.3700, 2.1693, 1.7052], + device='cuda:0'), covar=tensor([0.4942, 0.4761, 0.1774, 0.3683, 0.2352, 0.3022, 0.1760, 0.5535], + device='cuda:0'), in_proj_covar=tensor([0.0932, 0.0950, 0.0783, 0.0916, 0.0980, 0.0870, 0.0728, 0.0810], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 23:01:41,955 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=148860.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:01:53,950 INFO [optim.py:369] (0/4) 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 23:02:04,176 INFO [train.py:901] (0/4) Epoch 19, batch 3400, loss[loss=0.2201, simple_loss=0.3087, pruned_loss=0.06572, over 8446.00 frames. ], tot_loss[loss=0.2126, simple_loss=0.2928, pruned_loss=0.06619, over 1610093.05 frames. ], batch size: 29, lr: 4.00e-03, grad_scale: 4.0 +2023-02-06 23:02:13,143 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148906.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:02:38,059 INFO [train.py:901] (0/4) Epoch 19, batch 3450, loss[loss=0.2426, simple_loss=0.3202, pruned_loss=0.08249, over 8354.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.2927, pruned_loss=0.06601, over 1612544.24 frames. ], batch size: 26, lr: 4.00e-03, grad_scale: 4.0 +2023-02-06 23:03:01,937 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148975.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:03:04,405 INFO [optim.py:369] (0/4) 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 23:03:14,149 INFO [train.py:901] (0/4) Epoch 19, batch 3500, loss[loss=0.2536, simple_loss=0.3341, pruned_loss=0.08652, over 8462.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.2933, pruned_loss=0.06611, over 1614858.80 frames. ], batch size: 27, lr: 4.00e-03, grad_scale: 4.0 +2023-02-06 23:03:28,318 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=149013.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:03:33,358 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=149020.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:03:35,942 WARNING [train.py:1067] (0/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-06 23:03:48,899 INFO [train.py:901] (0/4) Epoch 19, batch 3550, loss[loss=0.2237, simple_loss=0.3052, pruned_loss=0.07107, over 8331.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.293, pruned_loss=0.06542, over 1615228.62 frames. ], batch size: 25, lr: 4.00e-03, grad_scale: 4.0 +2023-02-06 23:03:50,370 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=149045.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:04:13,072 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 23:04:25,646 INFO [train.py:901] (0/4) Epoch 19, batch 3600, loss[loss=0.1837, simple_loss=0.2656, pruned_loss=0.0509, over 7639.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2924, pruned_loss=0.065, over 1616682.95 frames. ], batch size: 19, lr: 4.00e-03, grad_scale: 8.0 +2023-02-06 23:04:36,706 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 +2023-02-06 23:04:49,816 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=149128.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:04:59,729 INFO [train.py:901] (0/4) Epoch 19, batch 3650, loss[loss=0.2374, simple_loss=0.3075, pruned_loss=0.08365, over 8042.00 frames. ], tot_loss[loss=0.2107, simple_loss=0.2917, pruned_loss=0.06488, over 1617395.05 frames. ], batch size: 22, lr: 4.00e-03, grad_scale: 8.0 +2023-02-06 23:05:13,237 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=149162.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:05:24,390 INFO [optim.py:369] (0/4) 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 23:05:30,729 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=149187.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:05:35,231 INFO [train.py:901] (0/4) Epoch 19, batch 3700, loss[loss=0.2392, simple_loss=0.3206, pruned_loss=0.07888, over 8691.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2918, pruned_loss=0.06501, over 1617385.55 frames. ], batch size: 34, lr: 4.00e-03, grad_scale: 8.0 +2023-02-06 23:05:35,405 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=149193.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:05:38,060 WARNING [train.py:1067] (0/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-06 23:06:02,787 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=149231.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:06:10,470 INFO [train.py:901] (0/4) Epoch 19, batch 3750, loss[loss=0.2264, simple_loss=0.2886, pruned_loss=0.08209, over 8245.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2914, pruned_loss=0.06514, over 1612457.05 frames. ], batch size: 22, lr: 4.00e-03, grad_scale: 8.0 +2023-02-06 23:06:19,366 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=149256.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:06:34,558 INFO [optim.py:369] (0/4) 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 23:06:44,223 INFO [train.py:901] (0/4) Epoch 19, batch 3800, loss[loss=0.1792, simple_loss=0.259, pruned_loss=0.04968, over 7934.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2908, pruned_loss=0.06508, over 1613206.60 frames. ], batch size: 20, lr: 4.00e-03, grad_scale: 8.0 +2023-02-06 23:07:20,712 INFO [train.py:901] (0/4) Epoch 19, batch 3850, loss[loss=0.2744, simple_loss=0.3452, pruned_loss=0.1018, over 7042.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2897, pruned_loss=0.06432, over 1612622.88 frames. ], batch size: 72, lr: 4.00e-03, grad_scale: 8.0 +2023-02-06 23:07:42,389 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-06 23:07:45,097 INFO [optim.py:369] (0/4) 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 23:07:48,732 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=149384.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:07:54,637 INFO [train.py:901] (0/4) Epoch 19, batch 3900, loss[loss=0.2053, simple_loss=0.283, pruned_loss=0.06381, over 8130.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2902, pruned_loss=0.06425, over 1615241.88 frames. ], batch size: 22, lr: 4.00e-03, grad_scale: 8.0 +2023-02-06 23:07:58,218 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7979, 1.5136, 2.8087, 1.3730, 2.1507, 2.9874, 3.1198, 2.5777], + device='cuda:0'), covar=tensor([0.1040, 0.1523, 0.0438, 0.2108, 0.0945, 0.0298, 0.0670, 0.0554], + device='cuda:0'), in_proj_covar=tensor([0.0285, 0.0313, 0.0281, 0.0308, 0.0298, 0.0261, 0.0400, 0.0300], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-06 23:07:58,927 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0333, 3.8501, 2.4207, 2.7561, 3.0043, 2.1516, 2.7917, 3.0199], + device='cuda:0'), covar=tensor([0.1698, 0.0342, 0.1071, 0.0840, 0.0737, 0.1319, 0.1170, 0.1228], + device='cuda:0'), in_proj_covar=tensor([0.0350, 0.0239, 0.0327, 0.0304, 0.0300, 0.0331, 0.0341, 0.0316], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 23:08:06,544 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=149409.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:08:31,938 INFO [train.py:901] (0/4) Epoch 19, batch 3950, loss[loss=0.1877, simple_loss=0.2672, pruned_loss=0.05414, over 6799.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.29, pruned_loss=0.0639, over 1613266.15 frames. ], batch size: 15, lr: 3.99e-03, grad_scale: 8.0 +2023-02-06 23:08:36,310 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=149449.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:08:53,035 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 23:09:04,519 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.6778, 4.6548, 4.2091, 2.0771, 4.0455, 4.2795, 4.1864, 3.9068], + device='cuda:0'), covar=tensor([0.0693, 0.0542, 0.1063, 0.4545, 0.0864, 0.0813, 0.1292, 0.0769], + device='cuda:0'), in_proj_covar=tensor([0.0522, 0.0430, 0.0432, 0.0533, 0.0422, 0.0437, 0.0415, 0.0377], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 23:09:05,731 INFO [train.py:901] (0/4) Epoch 19, batch 4000, loss[loss=0.1978, simple_loss=0.2846, pruned_loss=0.05547, over 8238.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2907, pruned_loss=0.06419, over 1612949.68 frames. ], batch size: 22, lr: 3.99e-03, grad_scale: 8.0 +2023-02-06 23:09:32,409 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 19, batch 4050, loss[loss=0.184, simple_loss=0.2537, pruned_loss=0.05708, over 7534.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2916, pruned_loss=0.06503, over 1614446.57 frames. ], batch size: 18, lr: 3.99e-03, grad_scale: 8.0 +2023-02-06 23:10:05,803 INFO [optim.py:369] (0/4) 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 23:10:08,583 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 19, batch 4100, loss[loss=0.2192, simple_loss=0.2861, pruned_loss=0.07617, over 7804.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.292, pruned_loss=0.06509, over 1616066.72 frames. ], batch size: 20, lr: 3.99e-03, grad_scale: 8.0 +2023-02-06 23:10:49,881 INFO [train.py:901] (0/4) Epoch 19, batch 4150, loss[loss=0.213, simple_loss=0.289, pruned_loss=0.06855, over 8341.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2919, pruned_loss=0.06468, over 1620611.80 frames. ], batch size: 26, lr: 3.99e-03, grad_scale: 8.0 +2023-02-06 23:10:57,882 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 +2023-02-06 23:11:16,655 INFO [optim.py:369] (0/4) 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 23:11:26,115 INFO [train.py:901] (0/4) Epoch 19, batch 4200, loss[loss=0.2267, simple_loss=0.3066, pruned_loss=0.07341, over 6955.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2926, pruned_loss=0.06541, over 1616455.95 frames. ], batch size: 71, lr: 3.99e-03, grad_scale: 8.0 +2023-02-06 23:11:36,583 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-06 23:11:41,377 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2120, 1.0723, 1.3025, 1.1076, 0.9888, 1.3318, 0.0675, 0.8825], + device='cuda:0'), covar=tensor([0.1747, 0.1400, 0.0550, 0.0935, 0.2963, 0.0571, 0.2424, 0.1349], + device='cuda:0'), in_proj_covar=tensor([0.0185, 0.0194, 0.0124, 0.0222, 0.0271, 0.0132, 0.0171, 0.0187], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 23:11:59,504 INFO [train.py:901] (0/4) Epoch 19, batch 4250, loss[loss=0.205, simple_loss=0.2916, pruned_loss=0.05917, over 8257.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2917, pruned_loss=0.06473, over 1617101.23 frames. ], batch size: 22, lr: 3.99e-03, grad_scale: 8.0 +2023-02-06 23:12:00,911 WARNING [train.py:1067] (0/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-06 23:12:01,074 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1219, 1.4188, 1.6223, 1.3045, 0.9049, 1.4767, 1.6853, 1.5894], + device='cuda:0'), covar=tensor([0.0535, 0.1321, 0.1711, 0.1471, 0.0641, 0.1520, 0.0726, 0.0651], + device='cuda:0'), in_proj_covar=tensor([0.0096, 0.0152, 0.0189, 0.0157, 0.0100, 0.0160, 0.0113, 0.0140], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 23:12:14,480 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=149764.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 23:12:25,317 INFO [optim.py:369] (0/4) 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 23:12:35,593 INFO [train.py:901] (0/4) Epoch 19, batch 4300, loss[loss=0.2137, simple_loss=0.315, pruned_loss=0.05622, over 8316.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2915, pruned_loss=0.06474, over 1616234.76 frames. ], batch size: 25, lr: 3.99e-03, grad_scale: 8.0 +2023-02-06 23:13:10,088 INFO [train.py:901] (0/4) Epoch 19, batch 4350, loss[loss=0.1926, simple_loss=0.2745, pruned_loss=0.05533, over 8241.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2914, pruned_loss=0.06507, over 1611651.93 frames. ], batch size: 22, lr: 3.99e-03, grad_scale: 8.0 +2023-02-06 23:13:33,155 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-06 23:13:33,211 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=149876.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:13:35,197 INFO [optim.py:369] (0/4) 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 23:13:44,579 INFO [train.py:901] (0/4) Epoch 19, batch 4400, loss[loss=0.1949, simple_loss=0.2743, pruned_loss=0.05772, over 8016.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2907, pruned_loss=0.06413, over 1611552.38 frames. ], batch size: 22, lr: 3.99e-03, grad_scale: 8.0 +2023-02-06 23:14:09,954 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=149927.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:14:14,588 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-06 23:14:20,890 INFO [train.py:901] (0/4) Epoch 19, batch 4450, loss[loss=0.1666, simple_loss=0.2499, pruned_loss=0.04162, over 7803.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2908, pruned_loss=0.06395, over 1607540.60 frames. ], batch size: 20, lr: 3.99e-03, grad_scale: 8.0 +2023-02-06 23:14:23,038 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6353, 1.4590, 1.7187, 1.3684, 0.7271, 1.4660, 1.4776, 1.4597], + device='cuda:0'), covar=tensor([0.0511, 0.1206, 0.1609, 0.1378, 0.0586, 0.1448, 0.0690, 0.0604], + device='cuda:0'), in_proj_covar=tensor([0.0096, 0.0151, 0.0188, 0.0158, 0.0100, 0.0161, 0.0112, 0.0140], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 23:14:44,886 INFO [optim.py:369] (0/4) 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 23:14:53,304 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=149991.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:14:54,477 INFO [train.py:901] (0/4) Epoch 19, batch 4500, loss[loss=0.21, simple_loss=0.3094, pruned_loss=0.05524, over 8318.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2914, pruned_loss=0.06433, over 1607210.41 frames. ], batch size: 25, lr: 3.99e-03, grad_scale: 8.0 +2023-02-06 23:14:59,956 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-150000.pt +2023-02-06 23:15:08,406 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-06 23:15:21,654 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3460, 1.4449, 1.3367, 1.7832, 0.8382, 1.2622, 1.2916, 1.4816], + device='cuda:0'), covar=tensor([0.0929, 0.0750, 0.1067, 0.0543, 0.1028, 0.1252, 0.0725, 0.0686], + device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0199, 0.0250, 0.0214, 0.0207, 0.0249, 0.0255, 0.0211], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 23:15:31,625 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150042.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:15:32,128 INFO [train.py:901] (0/4) Epoch 19, batch 4550, loss[loss=0.1927, simple_loss=0.2679, pruned_loss=0.05876, over 7421.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2915, pruned_loss=0.06414, over 1612145.51 frames. ], batch size: 17, lr: 3.99e-03, grad_scale: 8.0 +2023-02-06 23:15:54,569 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.1711, 1.0355, 1.2611, 1.0467, 0.9529, 1.2668, 0.1030, 0.9059], + device='cuda:0'), covar=tensor([0.1810, 0.1561, 0.0567, 0.0896, 0.3038, 0.0709, 0.2414, 0.1345], + device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0193, 0.0123, 0.0222, 0.0271, 0.0132, 0.0170, 0.0188], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 23:15:56,360 INFO [optim.py:369] (0/4) 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 23:16:05,765 INFO [train.py:901] (0/4) Epoch 19, batch 4600, loss[loss=0.2045, simple_loss=0.2952, pruned_loss=0.05691, over 8598.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2925, pruned_loss=0.06494, over 1610999.38 frames. ], batch size: 34, lr: 3.99e-03, grad_scale: 8.0 +2023-02-06 23:16:08,462 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=150097.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:16:15,961 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=150108.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 23:16:41,502 INFO [train.py:901] (0/4) Epoch 19, batch 4650, loss[loss=0.2226, simple_loss=0.3054, pruned_loss=0.06995, over 8506.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2915, pruned_loss=0.06449, over 1613926.16 frames. ], batch size: 26, lr: 3.99e-03, grad_scale: 8.0 +2023-02-06 23:17:03,546 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-02-06 23:17:06,561 INFO [optim.py:369] (0/4) 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 23:17:16,083 INFO [train.py:901] (0/4) Epoch 19, batch 4700, loss[loss=0.1738, simple_loss=0.2577, pruned_loss=0.045, over 7691.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2917, pruned_loss=0.06446, over 1614779.78 frames. ], batch size: 18, lr: 3.98e-03, grad_scale: 8.0 +2023-02-06 23:17:36,647 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150223.0, num_to_drop=1, layers_to_drop={1} +2023-02-06 23:17:39,458 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1232, 2.4552, 2.6217, 1.6246, 2.6941, 1.8473, 1.6660, 2.1082], + device='cuda:0'), covar=tensor([0.0680, 0.0348, 0.0238, 0.0657, 0.0431, 0.0722, 0.0735, 0.0443], + device='cuda:0'), in_proj_covar=tensor([0.0442, 0.0382, 0.0331, 0.0438, 0.0365, 0.0529, 0.0386, 0.0410], + device='cuda:0'), out_proj_covar=tensor([1.1973e-04, 1.0090e-04, 8.7497e-05, 1.1626e-04, 9.6710e-05, 1.5111e-04, + 1.0449e-04, 1.0962e-04], device='cuda:0') +2023-02-06 23:17:50,831 INFO [train.py:901] (0/4) Epoch 19, batch 4750, loss[loss=0.2073, simple_loss=0.2899, pruned_loss=0.06237, over 8028.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2912, pruned_loss=0.06393, over 1615350.09 frames. ], batch size: 22, lr: 3.98e-03, grad_scale: 8.0 +2023-02-06 23:17:53,777 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150247.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:18:12,328 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150272.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:18:12,661 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 +2023-02-06 23:18:13,491 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-06 23:18:15,508 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-06 23:18:16,855 INFO [optim.py:369] (0/4) 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 23:18:20,398 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.5602, 1.6927, 4.4269, 2.1210, 2.4560, 5.0518, 5.0996, 4.3712], + device='cuda:0'), covar=tensor([0.1090, 0.1836, 0.0278, 0.1815, 0.1147, 0.0169, 0.0426, 0.0547], + device='cuda:0'), in_proj_covar=tensor([0.0290, 0.0317, 0.0287, 0.0312, 0.0301, 0.0265, 0.0406, 0.0304], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 23:18:25,813 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=150292.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:18:26,362 INFO [train.py:901] (0/4) Epoch 19, batch 4800, loss[loss=0.2179, simple_loss=0.3102, pruned_loss=0.06273, over 8528.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2903, pruned_loss=0.06367, over 1611086.50 frames. ], batch size: 28, lr: 3.98e-03, grad_scale: 8.0 +2023-02-06 23:18:29,937 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150298.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:18:46,557 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150323.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:18:57,492 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.1989, 1.2908, 3.3782, 0.9789, 2.9364, 2.8132, 3.0679, 2.9556], + device='cuda:0'), covar=tensor([0.0920, 0.4045, 0.0832, 0.4274, 0.1485, 0.1169, 0.0761, 0.0976], + device='cuda:0'), in_proj_covar=tensor([0.0595, 0.0627, 0.0668, 0.0607, 0.0681, 0.0584, 0.0584, 0.0644], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 23:19:00,002 INFO [train.py:901] (0/4) Epoch 19, batch 4850, loss[loss=0.1787, simple_loss=0.2736, pruned_loss=0.04188, over 8305.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2917, pruned_loss=0.06453, over 1613817.62 frames. ], batch size: 23, lr: 3.98e-03, grad_scale: 8.0 +2023-02-06 23:19:05,338 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-06 23:19:27,011 INFO [optim.py:369] (0/4) 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 23:19:36,180 INFO [train.py:901] (0/4) Epoch 19, batch 4900, loss[loss=0.2077, simple_loss=0.2897, pruned_loss=0.06284, over 8619.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2902, pruned_loss=0.06403, over 1611677.63 frames. ], batch size: 31, lr: 3.98e-03, grad_scale: 8.0 +2023-02-06 23:20:06,481 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3023, 1.8465, 4.2709, 1.7116, 2.5539, 4.7893, 5.0683, 3.6963], + device='cuda:0'), covar=tensor([0.1478, 0.2030, 0.0373, 0.2625, 0.1204, 0.0321, 0.0502, 0.1021], + device='cuda:0'), in_proj_covar=tensor([0.0290, 0.0318, 0.0288, 0.0313, 0.0302, 0.0267, 0.0408, 0.0305], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 23:20:07,720 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=150441.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:20:08,994 INFO [train.py:901] (0/4) Epoch 19, batch 4950, loss[loss=0.1709, simple_loss=0.2565, pruned_loss=0.04267, over 7249.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2904, pruned_loss=0.06462, over 1607721.93 frames. ], batch size: 16, lr: 3.98e-03, grad_scale: 8.0 +2023-02-06 23:20:33,691 INFO [optim.py:369] (0/4) 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 23:20:33,932 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150479.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 23:20:43,976 INFO [train.py:901] (0/4) Epoch 19, batch 5000, loss[loss=0.1912, simple_loss=0.2821, pruned_loss=0.05016, over 8246.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2908, pruned_loss=0.06457, over 1609459.54 frames. ], batch size: 24, lr: 3.98e-03, grad_scale: 8.0 +2023-02-06 23:20:52,070 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150504.0, num_to_drop=1, layers_to_drop={0} +2023-02-06 23:21:06,772 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.7970, 1.4830, 3.9977, 1.4222, 3.4570, 3.2933, 3.5703, 3.4254], + device='cuda:0'), covar=tensor([0.0759, 0.4813, 0.0611, 0.4362, 0.1377, 0.1130, 0.0735, 0.0873], + device='cuda:0'), in_proj_covar=tensor([0.0598, 0.0632, 0.0671, 0.0610, 0.0684, 0.0589, 0.0590, 0.0649], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 23:21:15,546 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-02-06 23:21:17,792 INFO [train.py:901] (0/4) Epoch 19, batch 5050, loss[loss=0.2021, simple_loss=0.2855, pruned_loss=0.05933, over 8454.00 frames. ], tot_loss[loss=0.2107, simple_loss=0.2916, pruned_loss=0.06489, over 1613408.63 frames. ], batch size: 49, lr: 3.98e-03, grad_scale: 8.0 +2023-02-06 23:21:25,904 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=150555.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:21:26,606 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150556.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:21:31,777 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4095, 1.3168, 2.3786, 1.2269, 2.0897, 2.5194, 2.6799, 2.1472], + device='cuda:0'), covar=tensor([0.1161, 0.1442, 0.0458, 0.2173, 0.0717, 0.0407, 0.0650, 0.0742], + device='cuda:0'), in_proj_covar=tensor([0.0289, 0.0316, 0.0287, 0.0312, 0.0302, 0.0265, 0.0405, 0.0303], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 23:21:40,927 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-06 23:21:41,595 INFO [optim.py:369] (0/4) 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 23:21:51,768 INFO [train.py:901] (0/4) Epoch 19, batch 5100, loss[loss=0.1987, simple_loss=0.2828, pruned_loss=0.05727, over 7810.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2916, pruned_loss=0.06478, over 1612020.93 frames. ], batch size: 20, lr: 3.98e-03, grad_scale: 8.0 +2023-02-06 23:22:23,309 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=150636.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:22:27,858 INFO [train.py:901] (0/4) Epoch 19, batch 5150, loss[loss=0.2141, simple_loss=0.2964, pruned_loss=0.06587, over 8283.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2912, pruned_loss=0.06466, over 1612546.50 frames. ], batch size: 23, lr: 3.98e-03, grad_scale: 8.0 +2023-02-06 23:22:51,867 INFO [optim.py:369] (0/4) 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 23:22:53,379 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2226, 1.6320, 3.4114, 1.3764, 2.3141, 3.7606, 3.8714, 3.2188], + device='cuda:0'), covar=tensor([0.1062, 0.1714, 0.0347, 0.2263, 0.1057, 0.0234, 0.0512, 0.0570], + device='cuda:0'), in_proj_covar=tensor([0.0288, 0.0316, 0.0286, 0.0311, 0.0300, 0.0264, 0.0404, 0.0302], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-06 23:23:01,325 INFO [train.py:901] (0/4) Epoch 19, batch 5200, loss[loss=0.2255, simple_loss=0.3013, pruned_loss=0.07482, over 8447.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2897, pruned_loss=0.06386, over 1613754.90 frames. ], batch size: 49, lr: 3.98e-03, grad_scale: 8.0 +2023-02-06 23:23:04,484 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.24 vs. limit=5.0 +2023-02-06 23:23:38,104 INFO [train.py:901] (0/4) Epoch 19, batch 5250, loss[loss=0.2752, simple_loss=0.3377, pruned_loss=0.1063, over 6692.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.29, pruned_loss=0.06438, over 1608841.66 frames. ], batch size: 71, lr: 3.98e-03, grad_scale: 8.0 +2023-02-06 23:23:40,628 WARNING [train.py:1067] (0/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-06 23:23:42,647 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=150750.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:23:43,367 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150751.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:24:01,531 INFO [optim.py:369] (0/4) 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 23:24:10,883 INFO [train.py:901] (0/4) Epoch 19, batch 5300, loss[loss=0.2012, simple_loss=0.291, pruned_loss=0.05574, over 7821.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2908, pruned_loss=0.06485, over 1610157.02 frames. ], batch size: 20, lr: 3.98e-03, grad_scale: 8.0 +2023-02-06 23:24:21,007 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6326, 1.7357, 1.8007, 1.4307, 1.8810, 1.4735, 0.9551, 1.7466], + device='cuda:0'), covar=tensor([0.0403, 0.0323, 0.0183, 0.0363, 0.0318, 0.0544, 0.0622, 0.0196], + device='cuda:0'), in_proj_covar=tensor([0.0440, 0.0381, 0.0330, 0.0434, 0.0364, 0.0525, 0.0384, 0.0405], + device='cuda:0'), out_proj_covar=tensor([1.1926e-04, 1.0061e-04, 8.7090e-05, 1.1526e-04, 9.6612e-05, 1.4969e-04, + 1.0392e-04, 1.0818e-04], device='cuda:0') +2023-02-06 23:24:23,645 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150812.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:24:41,653 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150837.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:24:42,317 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7238, 1.8756, 1.6577, 2.3368, 1.1089, 1.4416, 1.7189, 1.9077], + device='cuda:0'), covar=tensor([0.0719, 0.0705, 0.0947, 0.0416, 0.0986, 0.1348, 0.0757, 0.0709], + device='cuda:0'), in_proj_covar=tensor([0.0230, 0.0197, 0.0247, 0.0212, 0.0203, 0.0248, 0.0252, 0.0209], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 23:24:46,243 INFO [train.py:901] (0/4) Epoch 19, batch 5350, loss[loss=0.2672, simple_loss=0.3386, pruned_loss=0.09786, over 8693.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.2928, pruned_loss=0.06613, over 1613748.86 frames. ], batch size: 39, lr: 3.98e-03, grad_scale: 8.0 +2023-02-06 23:25:10,958 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 19, batch 5400, loss[loss=0.1848, simple_loss=0.2668, pruned_loss=0.05145, over 7662.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2928, pruned_loss=0.06648, over 1616253.68 frames. ], batch size: 19, lr: 3.98e-03, grad_scale: 16.0 +2023-02-06 23:25:24,755 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=150899.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:25:37,919 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=150918.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:25:45,318 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.8682, 1.3130, 4.0096, 1.4453, 3.5462, 3.3726, 3.6140, 3.5043], + device='cuda:0'), covar=tensor([0.0616, 0.4712, 0.0623, 0.4148, 0.1224, 0.1007, 0.0659, 0.0756], + device='cuda:0'), in_proj_covar=tensor([0.0598, 0.0634, 0.0669, 0.0608, 0.0685, 0.0587, 0.0589, 0.0648], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 23:25:55,465 INFO [train.py:901] (0/4) Epoch 19, batch 5450, loss[loss=0.1992, simple_loss=0.2653, pruned_loss=0.06653, over 7436.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2911, pruned_loss=0.06541, over 1609797.04 frames. ], batch size: 17, lr: 3.97e-03, grad_scale: 16.0 +2023-02-06 23:25:55,706 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8345, 2.0136, 1.7444, 2.5623, 1.2511, 1.4906, 1.8048, 2.0659], + device='cuda:0'), covar=tensor([0.0736, 0.0709, 0.0905, 0.0396, 0.1045, 0.1324, 0.0787, 0.0677], + device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0197, 0.0247, 0.0212, 0.0204, 0.0247, 0.0252, 0.0209], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 23:26:22,623 INFO [optim.py:369] (0/4) 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 23:26:28,459 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 19, batch 5500, loss[loss=0.254, simple_loss=0.3224, pruned_loss=0.09278, over 7151.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2905, pruned_loss=0.06475, over 1611271.56 frames. ], batch size: 71, lr: 3.97e-03, grad_scale: 16.0 +2023-02-06 23:26:41,072 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.1808, 1.0404, 1.3053, 1.0708, 0.9556, 1.3271, 0.0986, 0.9232], + device='cuda:0'), covar=tensor([0.1700, 0.1446, 0.0507, 0.0879, 0.2800, 0.0597, 0.2277, 0.1345], + device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0193, 0.0123, 0.0222, 0.0269, 0.0132, 0.0169, 0.0188], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 23:26:41,788 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151007.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:26:46,613 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151014.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:26:58,950 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151032.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:27:00,495 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.48 vs. limit=2.0 +2023-02-06 23:27:06,205 INFO [train.py:901] (0/4) Epoch 19, batch 5550, loss[loss=0.2292, simple_loss=0.3074, pruned_loss=0.07552, over 8116.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.291, pruned_loss=0.06504, over 1613143.91 frames. ], batch size: 23, lr: 3.97e-03, grad_scale: 16.0 +2023-02-06 23:27:32,496 INFO [optim.py:369] (0/4) 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 23:27:42,410 INFO [train.py:901] (0/4) Epoch 19, batch 5600, loss[loss=0.1919, simple_loss=0.2805, pruned_loss=0.05167, over 8035.00 frames. ], tot_loss[loss=0.2116, simple_loss=0.2924, pruned_loss=0.06537, over 1616804.97 frames. ], batch size: 22, lr: 3.97e-03, grad_scale: 16.0 +2023-02-06 23:27:43,134 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=151094.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:28:15,318 INFO [train.py:901] (0/4) Epoch 19, batch 5650, loss[loss=0.2407, simple_loss=0.3266, pruned_loss=0.0774, over 8459.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2939, pruned_loss=0.06641, over 1619977.97 frames. ], batch size: 27, lr: 3.97e-03, grad_scale: 16.0 +2023-02-06 23:28:31,571 WARNING [train.py:1067] (0/4) 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] (0/4) 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 23:28:50,396 INFO [train.py:901] (0/4) Epoch 19, batch 5700, loss[loss=0.192, simple_loss=0.2759, pruned_loss=0.05407, over 7644.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2944, pruned_loss=0.067, over 1616587.20 frames. ], batch size: 19, lr: 3.97e-03, grad_scale: 16.0 +2023-02-06 23:28:54,075 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.7018, 2.3159, 3.3843, 1.7162, 1.7166, 3.3649, 0.5621, 1.9778], + device='cuda:0'), covar=tensor([0.1817, 0.1283, 0.0273, 0.2036, 0.3012, 0.0371, 0.2635, 0.1447], + device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0194, 0.0124, 0.0222, 0.0269, 0.0133, 0.0168, 0.0188], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 23:28:57,624 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151202.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:29:00,857 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.9391, 6.1429, 5.1961, 3.2964, 5.2948, 5.7404, 5.6050, 5.5754], + device='cuda:0'), covar=tensor([0.0521, 0.0361, 0.1028, 0.3426, 0.0655, 0.0723, 0.1143, 0.0562], + device='cuda:0'), in_proj_covar=tensor([0.0515, 0.0427, 0.0426, 0.0526, 0.0414, 0.0431, 0.0409, 0.0371], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 23:29:02,310 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151209.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:29:06,541 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-06 23:29:24,755 INFO [train.py:901] (0/4) Epoch 19, batch 5750, loss[loss=0.1629, simple_loss=0.2493, pruned_loss=0.03825, over 7925.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.2932, pruned_loss=0.06576, over 1618788.63 frames. ], batch size: 20, lr: 3.97e-03, grad_scale: 16.0 +2023-02-06 23:29:36,097 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-06 23:29:37,492 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=151262.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:29:43,023 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) Epoch 19, batch 5800, loss[loss=0.2187, simple_loss=0.296, pruned_loss=0.07074, over 7160.00 frames. ], tot_loss[loss=0.213, simple_loss=0.2935, pruned_loss=0.06622, over 1619701.63 frames. ], batch size: 71, lr: 3.97e-03, grad_scale: 16.0 +2023-02-06 23:30:00,348 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151295.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:30:04,371 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151300.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:30:16,584 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151316.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:30:20,001 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.1831, 1.2718, 4.3454, 1.5789, 3.8842, 3.6148, 3.9312, 3.8520], + device='cuda:0'), covar=tensor([0.0553, 0.4780, 0.0507, 0.4072, 0.1020, 0.0862, 0.0591, 0.0601], + device='cuda:0'), in_proj_covar=tensor([0.0600, 0.0634, 0.0673, 0.0608, 0.0685, 0.0586, 0.0590, 0.0651], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 23:30:24,044 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.1077, 1.5756, 4.3253, 1.6532, 3.7667, 3.5935, 3.8852, 3.7857], + device='cuda:0'), covar=tensor([0.0741, 0.4472, 0.0544, 0.3823, 0.1230, 0.0912, 0.0650, 0.0713], + device='cuda:0'), in_proj_covar=tensor([0.0601, 0.0635, 0.0674, 0.0608, 0.0686, 0.0587, 0.0591, 0.0651], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 23:30:34,584 INFO [train.py:901] (0/4) Epoch 19, batch 5850, loss[loss=0.188, simple_loss=0.262, pruned_loss=0.05694, over 7783.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.2929, pruned_loss=0.0659, over 1619749.84 frames. ], batch size: 19, lr: 3.97e-03, grad_scale: 16.0 +2023-02-06 23:30:57,557 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) Epoch 19, batch 5900, loss[loss=0.1994, simple_loss=0.2798, pruned_loss=0.05949, over 7247.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2927, pruned_loss=0.06572, over 1618224.72 frames. ], batch size: 16, lr: 3.97e-03, grad_scale: 16.0 +2023-02-06 23:31:21,367 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2800, 1.5364, 1.5691, 1.1199, 1.6211, 1.2632, 0.2933, 1.4800], + device='cuda:0'), covar=tensor([0.0533, 0.0432, 0.0386, 0.0562, 0.0544, 0.0999, 0.0969, 0.0319], + device='cuda:0'), in_proj_covar=tensor([0.0444, 0.0382, 0.0333, 0.0439, 0.0369, 0.0529, 0.0386, 0.0409], + device='cuda:0'), out_proj_covar=tensor([1.2022e-04, 1.0084e-04, 8.7799e-05, 1.1665e-04, 9.7842e-05, 1.5098e-04, + 1.0461e-04, 1.0906e-04], device='cuda:0') +2023-02-06 23:31:28,024 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6120, 1.9128, 2.1443, 1.4136, 2.2285, 1.3475, 0.6725, 1.8538], + device='cuda:0'), covar=tensor([0.0613, 0.0363, 0.0249, 0.0548, 0.0355, 0.1021, 0.0875, 0.0321], + device='cuda:0'), in_proj_covar=tensor([0.0444, 0.0381, 0.0333, 0.0439, 0.0368, 0.0528, 0.0386, 0.0408], + device='cuda:0'), out_proj_covar=tensor([1.2010e-04, 1.0074e-04, 8.7786e-05, 1.1660e-04, 9.7655e-05, 1.5087e-04, + 1.0459e-04, 1.0892e-04], device='cuda:0') +2023-02-06 23:31:37,730 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.63 vs. limit=5.0 +2023-02-06 23:31:44,903 INFO [train.py:901] (0/4) Epoch 19, batch 5950, loss[loss=0.2057, simple_loss=0.2905, pruned_loss=0.06041, over 8449.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2924, pruned_loss=0.06498, over 1619087.44 frames. ], batch size: 27, lr: 3.97e-03, grad_scale: 16.0 +2023-02-06 23:31:59,882 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 23:32:16,805 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151490.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:32:18,548 INFO [train.py:901] (0/4) Epoch 19, batch 6000, loss[loss=0.1851, simple_loss=0.2615, pruned_loss=0.05436, over 7295.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2928, pruned_loss=0.06573, over 1616521.25 frames. ], batch size: 16, lr: 3.97e-03, grad_scale: 16.0 +2023-02-06 23:32:18,549 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 23:32:32,013 INFO [train.py:935] (0/4) Epoch 19, validation: loss=0.1763, simple_loss=0.2764, pruned_loss=0.03805, over 944034.00 frames. +2023-02-06 23:32:32,014 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-06 23:33:06,937 INFO [train.py:901] (0/4) Epoch 19, batch 6050, loss[loss=0.1678, simple_loss=0.2568, pruned_loss=0.03937, over 8024.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.2922, pruned_loss=0.06538, over 1615781.19 frames. ], batch size: 22, lr: 3.97e-03, grad_scale: 16.0 +2023-02-06 23:33:09,100 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 23:33:42,766 INFO [train.py:901] (0/4) Epoch 19, batch 6100, loss[loss=0.2079, simple_loss=0.2908, pruned_loss=0.06249, over 8487.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.293, pruned_loss=0.06594, over 1614534.93 frames. ], batch size: 29, lr: 3.97e-03, grad_scale: 16.0 +2023-02-06 23:34:07,699 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-06 23:34:10,838 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151633.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:34:12,869 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.6864, 1.4790, 4.8984, 1.9093, 4.3970, 4.0936, 4.4341, 4.3531], + device='cuda:0'), covar=tensor([0.0511, 0.4460, 0.0438, 0.3789, 0.0876, 0.0866, 0.0500, 0.0539], + device='cuda:0'), in_proj_covar=tensor([0.0599, 0.0635, 0.0678, 0.0611, 0.0685, 0.0588, 0.0591, 0.0652], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 23:34:17,588 INFO [train.py:901] (0/4) Epoch 19, batch 6150, loss[loss=0.1944, simple_loss=0.2851, pruned_loss=0.05191, over 8682.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2929, pruned_loss=0.06556, over 1615255.74 frames. ], batch size: 34, lr: 3.97e-03, grad_scale: 16.0 +2023-02-06 23:34:18,358 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=151644.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:34:28,850 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151658.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:34:30,133 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=151660.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:34:30,967 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151661.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:34:43,583 INFO [optim.py:369] (0/4) 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 23:34:53,936 INFO [train.py:901] (0/4) Epoch 19, batch 6200, loss[loss=0.1792, simple_loss=0.2706, pruned_loss=0.0439, over 8086.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2912, pruned_loss=0.06469, over 1612446.69 frames. ], batch size: 21, lr: 3.96e-03, grad_scale: 16.0 +2023-02-06 23:35:02,690 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151706.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:35:10,161 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.12 vs. limit=5.0 +2023-02-06 23:35:28,542 INFO [train.py:901] (0/4) Epoch 19, batch 6250, loss[loss=0.1865, simple_loss=0.251, pruned_loss=0.06101, over 7192.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2911, pruned_loss=0.06455, over 1616353.33 frames. ], batch size: 16, lr: 3.96e-03, grad_scale: 16.0 +2023-02-06 23:35:39,357 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151759.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:35:50,918 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151775.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:35:53,500 INFO [optim.py:369] (0/4) 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 23:36:03,715 INFO [train.py:901] (0/4) Epoch 19, batch 6300, loss[loss=0.231, simple_loss=0.3056, pruned_loss=0.0782, over 8245.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2908, pruned_loss=0.06454, over 1613145.65 frames. ], batch size: 24, lr: 3.96e-03, grad_scale: 16.0 +2023-02-06 23:36:22,196 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151819.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:36:36,291 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 +2023-02-06 23:36:39,097 INFO [train.py:901] (0/4) Epoch 19, batch 6350, loss[loss=0.2258, simple_loss=0.3094, pruned_loss=0.07107, over 8494.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2904, pruned_loss=0.06455, over 1611344.21 frames. ], batch size: 29, lr: 3.96e-03, grad_scale: 16.0 +2023-02-06 23:36:52,340 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.4866, 1.2475, 3.8011, 1.5056, 2.9353, 2.9142, 3.2852, 3.3027], + device='cuda:0'), covar=tensor([0.1420, 0.7243, 0.1528, 0.5495, 0.2631, 0.2265, 0.1467, 0.1416], + device='cuda:0'), in_proj_covar=tensor([0.0592, 0.0627, 0.0667, 0.0603, 0.0679, 0.0583, 0.0584, 0.0641], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 23:37:03,138 INFO [optim.py:369] (0/4) 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 23:37:13,208 INFO [train.py:901] (0/4) Epoch 19, batch 6400, loss[loss=0.2086, simple_loss=0.2966, pruned_loss=0.06029, over 8512.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2915, pruned_loss=0.0652, over 1613561.24 frames. ], batch size: 29, lr: 3.96e-03, grad_scale: 16.0 +2023-02-06 23:37:17,661 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6354, 1.3980, 1.5961, 1.3166, 0.8619, 1.4400, 1.4637, 1.3583], + device='cuda:0'), covar=tensor([0.0586, 0.1276, 0.1717, 0.1462, 0.0632, 0.1498, 0.0714, 0.0660], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0152, 0.0190, 0.0159, 0.0100, 0.0161, 0.0112, 0.0142], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 23:37:30,618 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151917.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:37:48,044 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151942.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:37:48,519 INFO [train.py:901] (0/4) Epoch 19, batch 6450, loss[loss=0.1816, simple_loss=0.2707, pruned_loss=0.04618, over 8461.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2904, pruned_loss=0.06433, over 1614662.02 frames. ], batch size: 25, lr: 3.96e-03, grad_scale: 16.0 +2023-02-06 23:38:13,513 INFO [optim.py:369] (0/4) 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 23:38:20,582 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2263, 1.4109, 1.6631, 1.3277, 1.0407, 1.4184, 1.8612, 1.6820], + device='cuda:0'), covar=tensor([0.0494, 0.1385, 0.1712, 0.1519, 0.0616, 0.1584, 0.0690, 0.0653], + device='cuda:0'), in_proj_covar=tensor([0.0096, 0.0152, 0.0189, 0.0158, 0.0099, 0.0160, 0.0112, 0.0141], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0009, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 23:38:23,089 INFO [train.py:901] (0/4) Epoch 19, batch 6500, loss[loss=0.2065, simple_loss=0.2851, pruned_loss=0.06397, over 8438.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2899, pruned_loss=0.06427, over 1614866.27 frames. ], batch size: 27, lr: 3.96e-03, grad_scale: 16.0 +2023-02-06 23:38:27,747 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-152000.pt +2023-02-06 23:38:40,007 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152015.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:38:51,607 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152040.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:39:00,452 INFO [train.py:901] (0/4) Epoch 19, batch 6550, loss[loss=0.1712, simple_loss=0.2509, pruned_loss=0.0458, over 7698.00 frames. ], tot_loss[loss=0.208, simple_loss=0.289, pruned_loss=0.06347, over 1613098.80 frames. ], batch size: 18, lr: 3.96e-03, grad_scale: 16.0 +2023-02-06 23:39:04,922 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=152050.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:39:09,220 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152056.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:39:21,603 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-06 23:39:24,915 INFO [optim.py:369] (0/4) 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 23:39:34,324 INFO [train.py:901] (0/4) Epoch 19, batch 6600, loss[loss=0.2557, simple_loss=0.3261, pruned_loss=0.09261, over 8634.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2902, pruned_loss=0.06391, over 1616530.33 frames. ], batch size: 34, lr: 3.96e-03, grad_scale: 16.0 +2023-02-06 23:39:39,624 WARNING [train.py:1067] (0/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-06 23:40:09,003 INFO [train.py:901] (0/4) Epoch 19, batch 6650, loss[loss=0.2328, simple_loss=0.3176, pruned_loss=0.07398, over 8760.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2904, pruned_loss=0.06403, over 1614765.92 frames. ], batch size: 30, lr: 3.96e-03, grad_scale: 16.0 +2023-02-06 23:40:23,461 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=152163.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:40:24,942 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=152165.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:40:34,177 INFO [optim.py:369] (0/4) 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 23:40:44,520 INFO [train.py:901] (0/4) Epoch 19, batch 6700, loss[loss=0.1637, simple_loss=0.2346, pruned_loss=0.04644, over 6785.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2899, pruned_loss=0.0635, over 1614739.55 frames. ], batch size: 15, lr: 3.96e-03, grad_scale: 16.0 +2023-02-06 23:40:46,809 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0845, 2.1868, 1.8681, 2.7789, 1.3423, 1.6151, 2.0062, 2.3547], + device='cuda:0'), covar=tensor([0.0735, 0.0809, 0.0907, 0.0419, 0.1099, 0.1396, 0.0868, 0.0698], + device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0198, 0.0249, 0.0213, 0.0206, 0.0250, 0.0254, 0.0210], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 23:41:01,636 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4020, 1.6864, 1.6668, 1.0381, 1.7440, 1.3710, 0.2503, 1.6350], + device='cuda:0'), covar=tensor([0.0432, 0.0317, 0.0309, 0.0471, 0.0382, 0.0888, 0.0769, 0.0240], + device='cuda:0'), in_proj_covar=tensor([0.0438, 0.0377, 0.0330, 0.0436, 0.0363, 0.0521, 0.0382, 0.0403], + device='cuda:0'), out_proj_covar=tensor([1.1851e-04, 9.9523e-05, 8.7250e-05, 1.1574e-04, 9.6159e-05, 1.4864e-04, + 1.0355e-04, 1.0741e-04], device='cuda:0') +2023-02-06 23:41:19,456 INFO [train.py:901] (0/4) Epoch 19, batch 6750, loss[loss=0.1971, simple_loss=0.2785, pruned_loss=0.05783, over 8237.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2912, pruned_loss=0.06433, over 1614707.24 frames. ], batch size: 22, lr: 3.96e-03, grad_scale: 16.0 +2023-02-06 23:41:19,594 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152243.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:41:40,040 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.51 vs. limit=2.0 +2023-02-06 23:41:41,437 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 +2023-02-06 23:41:44,675 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 23:41:53,495 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152291.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:41:54,063 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-06 23:41:54,748 INFO [train.py:901] (0/4) Epoch 19, batch 6800, loss[loss=0.2337, simple_loss=0.3117, pruned_loss=0.07779, over 8472.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.291, pruned_loss=0.06402, over 1615735.05 frames. ], batch size: 27, lr: 3.96e-03, grad_scale: 16.0 +2023-02-06 23:42:29,090 INFO [train.py:901] (0/4) Epoch 19, batch 6850, loss[loss=0.1958, simple_loss=0.2876, pruned_loss=0.05196, over 8328.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2903, pruned_loss=0.0638, over 1612219.44 frames. ], batch size: 25, lr: 3.96e-03, grad_scale: 16.0 +2023-02-06 23:42:43,965 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-06 23:42:54,741 INFO [optim.py:369] (0/4) 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 23:43:05,118 INFO [train.py:901] (0/4) Epoch 19, batch 6900, loss[loss=0.197, simple_loss=0.2865, pruned_loss=0.05378, over 8290.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2905, pruned_loss=0.06359, over 1613091.94 frames. ], batch size: 23, lr: 3.96e-03, grad_scale: 16.0 +2023-02-06 23:43:10,945 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9003, 1.6508, 2.0456, 1.7569, 1.9278, 1.9336, 1.7140, 0.7849], + device='cuda:0'), covar=tensor([0.5021, 0.4359, 0.1826, 0.3187, 0.2336, 0.2720, 0.1915, 0.4580], + device='cuda:0'), in_proj_covar=tensor([0.0927, 0.0956, 0.0783, 0.0918, 0.0977, 0.0872, 0.0738, 0.0812], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 23:43:25,518 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152421.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:43:29,729 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7199, 2.0754, 2.1952, 1.3839, 2.2712, 1.5467, 0.6487, 1.9349], + device='cuda:0'), covar=tensor([0.0477, 0.0294, 0.0260, 0.0534, 0.0349, 0.0819, 0.0757, 0.0283], + device='cuda:0'), in_proj_covar=tensor([0.0437, 0.0378, 0.0333, 0.0437, 0.0365, 0.0524, 0.0384, 0.0406], + device='cuda:0'), out_proj_covar=tensor([1.1828e-04, 9.9567e-05, 8.8043e-05, 1.1599e-04, 9.6806e-05, 1.4952e-04, + 1.0402e-04, 1.0828e-04], device='cuda:0') +2023-02-06 23:43:40,384 INFO [train.py:901] (0/4) Epoch 19, batch 6950, loss[loss=0.2364, simple_loss=0.3263, pruned_loss=0.07323, over 8094.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2919, pruned_loss=0.06439, over 1616169.68 frames. ], batch size: 23, lr: 3.96e-03, grad_scale: 16.0 +2023-02-06 23:43:42,596 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152446.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:43:53,786 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-06 23:43:53,933 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152463.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:44:05,258 INFO [optim.py:369] (0/4) 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 23:44:14,623 INFO [train.py:901] (0/4) Epoch 19, batch 7000, loss[loss=0.2035, simple_loss=0.2851, pruned_loss=0.06093, over 8133.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2929, pruned_loss=0.06479, over 1620069.50 frames. ], batch size: 22, lr: 3.95e-03, grad_scale: 16.0 +2023-02-06 23:44:44,356 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152534.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:44:51,103 INFO [train.py:901] (0/4) Epoch 19, batch 7050, loss[loss=0.2101, simple_loss=0.2882, pruned_loss=0.066, over 7961.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.293, pruned_loss=0.06477, over 1619451.21 frames. ], batch size: 21, lr: 3.95e-03, grad_scale: 16.0 +2023-02-06 23:45:02,279 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152559.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:45:15,723 INFO [optim.py:369] (0/4) 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 23:45:21,283 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=152587.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:45:25,398 INFO [train.py:901] (0/4) Epoch 19, batch 7100, loss[loss=0.2448, simple_loss=0.3211, pruned_loss=0.08427, over 8569.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2925, pruned_loss=0.0649, over 1619922.13 frames. ], batch size: 31, lr: 3.95e-03, grad_scale: 16.0 +2023-02-06 23:45:30,673 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152600.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:45:35,438 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152607.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:45:56,366 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=152635.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:46:00,031 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1218, 1.9135, 2.5473, 2.0367, 2.3394, 2.1532, 1.8618, 1.2238], + device='cuda:0'), covar=tensor([0.5144, 0.4450, 0.1573, 0.3078, 0.2304, 0.2827, 0.1889, 0.4674], + device='cuda:0'), in_proj_covar=tensor([0.0929, 0.0957, 0.0783, 0.0921, 0.0979, 0.0872, 0.0738, 0.0816], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 23:46:01,878 INFO [train.py:901] (0/4) Epoch 19, batch 7150, loss[loss=0.1923, simple_loss=0.2698, pruned_loss=0.05739, over 7810.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.292, pruned_loss=0.06451, over 1620153.12 frames. ], batch size: 20, lr: 3.95e-03, grad_scale: 16.0 +2023-02-06 23:46:27,177 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 19, batch 7200, loss[loss=0.2267, simple_loss=0.299, pruned_loss=0.07716, over 8211.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2912, pruned_loss=0.06464, over 1616361.06 frames. ], batch size: 23, lr: 3.95e-03, grad_scale: 16.0 +2023-02-06 23:46:42,736 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=152702.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:46:44,095 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6389, 1.6697, 2.0709, 1.4437, 1.2214, 2.0501, 0.3236, 1.3199], + device='cuda:0'), covar=tensor([0.1828, 0.1353, 0.0418, 0.1196, 0.2931, 0.0459, 0.2170, 0.1337], + device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0194, 0.0123, 0.0221, 0.0267, 0.0133, 0.0168, 0.0186], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 23:47:12,601 INFO [train.py:901] (0/4) Epoch 19, batch 7250, loss[loss=0.2292, simple_loss=0.3124, pruned_loss=0.073, over 8249.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2907, pruned_loss=0.06469, over 1615255.90 frames. ], batch size: 24, lr: 3.95e-03, grad_scale: 16.0 +2023-02-06 23:47:13,459 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152744.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:47:17,250 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 +2023-02-06 23:47:17,701 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=152750.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:47:37,385 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 19, batch 7300, loss[loss=0.1788, simple_loss=0.2581, pruned_loss=0.04976, over 7422.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2903, pruned_loss=0.06453, over 1611880.15 frames. ], batch size: 17, lr: 3.95e-03, grad_scale: 16.0 +2023-02-06 23:47:57,325 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=152807.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:48:21,885 INFO [train.py:901] (0/4) Epoch 19, batch 7350, loss[loss=0.2541, simple_loss=0.3262, pruned_loss=0.09105, over 7929.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2902, pruned_loss=0.06476, over 1610776.97 frames. ], batch size: 20, lr: 3.95e-03, grad_scale: 16.0 +2023-02-06 23:48:41,691 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.56 vs. limit=5.0 +2023-02-06 23:48:46,710 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-06 23:48:48,162 INFO [optim.py:369] (0/4) 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 23:48:58,050 INFO [train.py:901] (0/4) Epoch 19, batch 7400, loss[loss=0.2255, simple_loss=0.3047, pruned_loss=0.0731, over 8582.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2905, pruned_loss=0.06442, over 1615362.90 frames. ], batch size: 39, lr: 3.95e-03, grad_scale: 32.0 +2023-02-06 23:49:07,695 WARNING [train.py:1067] (0/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-06 23:49:13,379 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3079, 2.0046, 4.4023, 2.1052, 2.4352, 5.0346, 5.1102, 4.3487], + device='cuda:0'), covar=tensor([0.1227, 0.1581, 0.0278, 0.1904, 0.1224, 0.0176, 0.0396, 0.0569], + device='cuda:0'), in_proj_covar=tensor([0.0292, 0.0320, 0.0286, 0.0313, 0.0303, 0.0262, 0.0409, 0.0306], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 23:49:18,798 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=152922.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:49:32,918 INFO [train.py:901] (0/4) Epoch 19, batch 7450, loss[loss=0.1719, simple_loss=0.2558, pruned_loss=0.04404, over 7556.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2905, pruned_loss=0.06425, over 1613918.17 frames. ], batch size: 18, lr: 3.95e-03, grad_scale: 32.0 +2023-02-06 23:49:33,696 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=152944.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:49:38,517 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=152951.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:49:44,014 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152958.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:49:46,572 WARNING [train.py:1067] (0/4) 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] (0/4) 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 23:50:01,203 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 19, batch 7500, loss[loss=0.213, simple_loss=0.2832, pruned_loss=0.07136, over 7264.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2904, pruned_loss=0.06452, over 1609356.96 frames. ], batch size: 16, lr: 3.95e-03, grad_scale: 16.0 +2023-02-06 23:50:17,495 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153006.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:50:31,548 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5340, 2.5920, 1.9645, 2.2708, 2.1451, 1.6126, 2.0776, 2.1848], + device='cuda:0'), covar=tensor([0.1388, 0.0385, 0.1027, 0.0623, 0.0697, 0.1448, 0.0870, 0.0919], + device='cuda:0'), in_proj_covar=tensor([0.0345, 0.0232, 0.0324, 0.0300, 0.0295, 0.0327, 0.0338, 0.0309], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 23:50:34,833 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=153031.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:50:42,929 INFO [train.py:901] (0/4) Epoch 19, batch 7550, loss[loss=0.2271, simple_loss=0.3024, pruned_loss=0.07594, over 8029.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2905, pruned_loss=0.06414, over 1614112.67 frames. ], batch size: 22, lr: 3.95e-03, grad_scale: 16.0 +2023-02-06 23:50:47,919 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.8630, 3.0895, 2.5378, 4.1577, 1.8722, 2.2263, 2.7231, 3.3364], + device='cuda:0'), covar=tensor([0.0638, 0.0806, 0.0805, 0.0240, 0.1058, 0.1191, 0.0893, 0.0724], + device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0196, 0.0246, 0.0211, 0.0203, 0.0246, 0.0250, 0.0207], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-06 23:50:53,917 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153059.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:50:58,707 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153066.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:50:58,948 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 +2023-02-06 23:51:08,489 INFO [optim.py:369] (0/4) 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 23:51:14,080 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 19, batch 7600, loss[loss=0.2299, simple_loss=0.3136, pruned_loss=0.07315, over 8486.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2912, pruned_loss=0.06472, over 1611388.31 frames. ], batch size: 28, lr: 3.95e-03, grad_scale: 16.0 +2023-02-06 23:51:45,038 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2467, 1.1776, 1.4078, 1.1237, 0.7545, 1.2099, 1.2389, 1.0336], + device='cuda:0'), covar=tensor([0.0619, 0.1245, 0.1647, 0.1458, 0.0606, 0.1458, 0.0698, 0.0695], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0152, 0.0190, 0.0158, 0.0100, 0.0160, 0.0113, 0.0142], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 23:51:47,761 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=153135.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:51:53,093 INFO [train.py:901] (0/4) Epoch 19, batch 7650, loss[loss=0.2182, simple_loss=0.296, pruned_loss=0.07023, over 8251.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2901, pruned_loss=0.06452, over 1611340.80 frames. ], batch size: 24, lr: 3.95e-03, grad_scale: 16.0 +2023-02-06 23:52:17,637 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 23:52:28,387 INFO [train.py:901] (0/4) Epoch 19, batch 7700, loss[loss=0.1792, simple_loss=0.2626, pruned_loss=0.04785, over 7551.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2904, pruned_loss=0.06452, over 1611231.69 frames. ], batch size: 18, lr: 3.95e-03, grad_scale: 16.0 +2023-02-06 23:52:33,499 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3822, 1.9911, 4.1499, 1.3207, 2.7797, 1.9927, 1.4392, 2.5791], + device='cuda:0'), covar=tensor([0.2268, 0.2930, 0.0763, 0.5097, 0.2115, 0.3480, 0.2727, 0.2756], + device='cuda:0'), in_proj_covar=tensor([0.0519, 0.0587, 0.0557, 0.0632, 0.0644, 0.0592, 0.0524, 0.0635], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 23:52:35,486 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=153203.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:52:44,963 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.2547, 1.1442, 3.3696, 0.9920, 2.9856, 2.8022, 3.0718, 2.9812], + device='cuda:0'), covar=tensor([0.0805, 0.4550, 0.0798, 0.4359, 0.1409, 0.1168, 0.0759, 0.0908], + device='cuda:0'), in_proj_covar=tensor([0.0600, 0.0630, 0.0673, 0.0605, 0.0687, 0.0589, 0.0586, 0.0647], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 23:52:57,462 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-06 23:53:02,813 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.2536, 1.1589, 3.3633, 0.9961, 2.9553, 2.8176, 3.0605, 2.9532], + device='cuda:0'), covar=tensor([0.0820, 0.4305, 0.0858, 0.4317, 0.1410, 0.1109, 0.0748, 0.0918], + device='cuda:0'), in_proj_covar=tensor([0.0596, 0.0627, 0.0669, 0.0602, 0.0684, 0.0586, 0.0584, 0.0644], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 23:53:03,343 INFO [train.py:901] (0/4) Epoch 19, batch 7750, loss[loss=0.2592, simple_loss=0.3345, pruned_loss=0.09194, over 8515.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2922, pruned_loss=0.06488, over 1617876.88 frames. ], batch size: 26, lr: 3.94e-03, grad_scale: 16.0 +2023-02-06 23:53:15,051 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 +2023-02-06 23:53:18,415 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.2415, 4.2108, 3.8228, 2.1936, 3.7574, 3.8388, 3.8264, 3.5857], + device='cuda:0'), covar=tensor([0.0762, 0.0585, 0.1168, 0.4332, 0.0930, 0.0885, 0.1362, 0.0860], + device='cuda:0'), in_proj_covar=tensor([0.0522, 0.0429, 0.0429, 0.0536, 0.0422, 0.0434, 0.0416, 0.0375], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 23:53:28,921 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 19, batch 7800, loss[loss=0.2368, simple_loss=0.3242, pruned_loss=0.07472, over 8504.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.2924, pruned_loss=0.06534, over 1613693.55 frames. ], batch size: 26, lr: 3.94e-03, grad_scale: 16.0 +2023-02-06 23:53:39,913 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=153296.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:53:53,366 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153315.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:53:57,897 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.9564, 3.7534, 3.4499, 2.5619, 3.4083, 3.4912, 3.5667, 3.2897], + device='cuda:0'), covar=tensor([0.0736, 0.0771, 0.1092, 0.3508, 0.0826, 0.1179, 0.1248, 0.1022], + device='cuda:0'), in_proj_covar=tensor([0.0524, 0.0430, 0.0431, 0.0537, 0.0423, 0.0435, 0.0417, 0.0375], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 23:53:58,012 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153322.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:54:06,990 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6618, 2.1828, 4.0877, 1.4488, 3.0102, 2.1876, 1.7307, 2.9959], + device='cuda:0'), covar=tensor([0.1902, 0.2666, 0.0795, 0.4536, 0.1860, 0.3141, 0.2351, 0.2297], + device='cuda:0'), in_proj_covar=tensor([0.0516, 0.0584, 0.0555, 0.0629, 0.0641, 0.0589, 0.0522, 0.0631], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-06 23:54:09,673 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 19, batch 7850, loss[loss=0.1838, simple_loss=0.2671, pruned_loss=0.0503, over 7945.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2926, pruned_loss=0.06583, over 1608927.29 frames. ], batch size: 20, lr: 3.94e-03, grad_scale: 8.0 +2023-02-06 23:54:14,371 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=153347.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:54:36,623 INFO [optim.py:369] (0/4) 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 23:54:44,304 INFO [train.py:901] (0/4) Epoch 19, batch 7900, loss[loss=0.2354, simple_loss=0.3232, pruned_loss=0.07384, over 8488.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2902, pruned_loss=0.06452, over 1604882.91 frames. ], batch size: 29, lr: 3.94e-03, grad_scale: 8.0 +2023-02-06 23:55:15,521 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 19, batch 7950, loss[loss=0.2561, simple_loss=0.3246, pruned_loss=0.09386, over 7096.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2907, pruned_loss=0.06449, over 1607443.04 frames. ], batch size: 71, lr: 3.94e-03, grad_scale: 8.0 +2023-02-06 23:55:28,805 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153459.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:55:37,525 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3447, 1.9488, 2.5490, 2.0919, 2.4640, 2.3329, 2.1518, 1.3328], + device='cuda:0'), covar=tensor([0.5001, 0.4694, 0.1913, 0.3811, 0.2496, 0.2731, 0.1662, 0.5079], + device='cuda:0'), in_proj_covar=tensor([0.0926, 0.0955, 0.0782, 0.0922, 0.0981, 0.0867, 0.0735, 0.0809], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-06 23:55:41,874 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 23:55:45,334 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=153484.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:55:51,001 INFO [train.py:901] (0/4) Epoch 19, batch 8000, loss[loss=0.2342, simple_loss=0.332, pruned_loss=0.06817, over 8340.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.291, pruned_loss=0.06485, over 1607962.26 frames. ], batch size: 25, lr: 3.94e-03, grad_scale: 8.0 +2023-02-06 23:56:17,602 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-06 23:56:25,144 INFO [train.py:901] (0/4) Epoch 19, batch 8050, loss[loss=0.1901, simple_loss=0.2707, pruned_loss=0.05477, over 7519.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2896, pruned_loss=0.06424, over 1602488.85 frames. ], batch size: 18, lr: 3.94e-03, grad_scale: 8.0 +2023-02-06 23:56:48,472 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/epoch-19.pt +2023-02-06 23:57:01,097 WARNING [train.py:1067] (0/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-06 23:57:04,930 INFO [train.py:901] (0/4) Epoch 20, batch 0, loss[loss=0.2063, simple_loss=0.2879, pruned_loss=0.06231, over 8075.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2879, pruned_loss=0.06231, over 8075.00 frames. ], batch size: 21, lr: 3.84e-03, grad_scale: 8.0 +2023-02-06 23:57:04,931 INFO [train.py:926] (0/4) Computing validation loss +2023-02-06 23:57:16,945 INFO [train.py:935] (0/4) 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] (0/4) Maximum memory allocated so far is 6722MB +2023-02-06 23:57:20,459 INFO [optim.py:369] (0/4) 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 23:57:29,439 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6436, 1.2447, 1.5287, 1.2110, 0.8722, 1.2821, 1.5529, 1.5428], + device='cuda:0'), covar=tensor([0.0611, 0.1841, 0.2539, 0.1964, 0.0696, 0.2110, 0.0822, 0.0733], + device='cuda:0'), in_proj_covar=tensor([0.0098, 0.0153, 0.0191, 0.0158, 0.0100, 0.0161, 0.0113, 0.0143], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-06 23:57:29,447 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153594.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:57:31,306 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-06 23:57:33,665 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2838, 2.0904, 3.1185, 1.9157, 2.6341, 3.4413, 3.3388, 3.1023], + device='cuda:0'), covar=tensor([0.0944, 0.1325, 0.0552, 0.1682, 0.1200, 0.0226, 0.0628, 0.0437], + device='cuda:0'), in_proj_covar=tensor([0.0288, 0.0317, 0.0284, 0.0310, 0.0300, 0.0260, 0.0403, 0.0303], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-06 23:57:41,242 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3395, 2.4129, 1.7375, 2.0638, 2.0123, 1.4812, 1.7514, 1.8841], + device='cuda:0'), covar=tensor([0.1336, 0.0368, 0.1078, 0.0611, 0.0650, 0.1419, 0.0950, 0.0873], + device='cuda:0'), in_proj_covar=tensor([0.0352, 0.0235, 0.0330, 0.0305, 0.0299, 0.0332, 0.0343, 0.0314], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-06 23:57:51,326 INFO [train.py:901] (0/4) Epoch 20, batch 50, loss[loss=0.1908, simple_loss=0.2822, pruned_loss=0.04968, over 8310.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.2957, pruned_loss=0.06669, over 368384.43 frames. ], batch size: 25, lr: 3.84e-03, grad_scale: 8.0 +2023-02-06 23:58:01,101 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=153640.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:58:06,570 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 20, batch 100, loss[loss=0.2435, simple_loss=0.3187, pruned_loss=0.08412, over 8465.00 frames. ], tot_loss[loss=0.213, simple_loss=0.2938, pruned_loss=0.06616, over 651407.07 frames. ], batch size: 29, lr: 3.84e-03, grad_scale: 8.0 +2023-02-06 23:58:29,238 WARNING [train.py:1067] (0/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-06 23:58:31,357 INFO [optim.py:369] (0/4) 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 23:59:03,135 INFO [train.py:901] (0/4) Epoch 20, batch 150, loss[loss=0.1716, simple_loss=0.26, pruned_loss=0.04164, over 8088.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2924, pruned_loss=0.06491, over 863181.44 frames. ], batch size: 21, lr: 3.84e-03, grad_scale: 8.0 +2023-02-06 23:59:23,316 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153755.0, num_to_drop=0, layers_to_drop=set() +2023-02-06 23:59:26,114 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7845, 1.8053, 2.3207, 1.6076, 1.3618, 2.3303, 0.5246, 1.4482], + device='cuda:0'), covar=tensor([0.1728, 0.1183, 0.0303, 0.1178, 0.2819, 0.0388, 0.2148, 0.1302], + device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0194, 0.0124, 0.0221, 0.0270, 0.0133, 0.0169, 0.0186], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-06 23:59:39,298 INFO [train.py:901] (0/4) Epoch 20, batch 200, loss[loss=0.1828, simple_loss=0.2599, pruned_loss=0.05284, over 7555.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2918, pruned_loss=0.06498, over 1027477.45 frames. ], batch size: 18, lr: 3.84e-03, grad_scale: 8.0 +2023-02-06 23:59:42,412 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.10 vs. limit=5.0 +2023-02-06 23:59:42,527 INFO [optim.py:369] (0/4) 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 23:59:43,904 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=153783.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:00:15,038 INFO [train.py:901] (0/4) Epoch 20, batch 250, loss[loss=0.1665, simple_loss=0.2449, pruned_loss=0.04406, over 7703.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2919, pruned_loss=0.06537, over 1151943.89 frames. ], batch size: 18, lr: 3.84e-03, grad_scale: 8.0 +2023-02-07 00:00:26,531 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-07 00:00:31,606 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153850.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:00:34,729 WARNING [train.py:1067] (0/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-07 00:00:48,269 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=153875.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:00:48,764 INFO [train.py:901] (0/4) Epoch 20, batch 300, loss[loss=0.1997, simple_loss=0.2775, pruned_loss=0.06096, over 7697.00 frames. ], tot_loss[loss=0.213, simple_loss=0.2936, pruned_loss=0.06617, over 1256625.10 frames. ], batch size: 18, lr: 3.84e-03, grad_scale: 8.0 +2023-02-07 00:00:51,992 INFO [optim.py:369] (0/4) 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 00:01:05,165 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153898.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:01:24,553 INFO [train.py:901] (0/4) Epoch 20, batch 350, loss[loss=0.2184, simple_loss=0.3032, pruned_loss=0.06685, over 8029.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2924, pruned_loss=0.06553, over 1337805.14 frames. ], batch size: 22, lr: 3.83e-03, grad_scale: 8.0 +2023-02-07 00:01:35,765 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=153941.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:01:36,640 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 +2023-02-07 00:01:59,284 INFO [train.py:901] (0/4) Epoch 20, batch 400, loss[loss=0.1663, simple_loss=0.2466, pruned_loss=0.04298, over 7698.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.2934, pruned_loss=0.0661, over 1399324.47 frames. ], batch size: 18, lr: 3.83e-03, grad_scale: 8.0 +2023-02-07 00:02:02,806 INFO [optim.py:369] (0/4) 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 00:02:15,726 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-154000.pt +2023-02-07 00:02:22,964 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-02-07 00:02:25,630 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 20, batch 450, loss[loss=0.2205, simple_loss=0.3032, pruned_loss=0.06888, over 8743.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.2926, pruned_loss=0.06519, over 1449912.76 frames. ], batch size: 30, lr: 3.83e-03, grad_scale: 8.0 +2023-02-07 00:02:44,055 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154036.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:03:05,762 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154067.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 00:03:11,785 INFO [train.py:901] (0/4) Epoch 20, batch 500, loss[loss=0.1899, simple_loss=0.2832, pruned_loss=0.04829, over 8641.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2907, pruned_loss=0.06427, over 1482364.45 frames. ], batch size: 49, lr: 3.83e-03, grad_scale: 8.0 +2023-02-07 00:03:15,237 INFO [optim.py:369] (0/4) 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 00:03:24,422 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154094.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:03:29,957 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154102.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:03:46,381 INFO [train.py:901] (0/4) Epoch 20, batch 550, loss[loss=0.1458, simple_loss=0.2215, pruned_loss=0.03505, over 7430.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2904, pruned_loss=0.06376, over 1509865.49 frames. ], batch size: 17, lr: 3.83e-03, grad_scale: 8.0 +2023-02-07 00:03:49,725 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 +2023-02-07 00:03:59,234 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.59 vs. limit=2.0 +2023-02-07 00:04:07,825 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154154.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:04:23,276 INFO [train.py:901] (0/4) Epoch 20, batch 600, loss[loss=0.2074, simple_loss=0.2775, pruned_loss=0.06866, over 7979.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2892, pruned_loss=0.06324, over 1530989.74 frames. ], batch size: 21, lr: 3.83e-03, grad_scale: 8.0 +2023-02-07 00:04:25,537 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154179.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:04:26,651 INFO [optim.py:369] (0/4) 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 00:04:45,282 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-07 00:04:57,538 INFO [train.py:901] (0/4) Epoch 20, batch 650, loss[loss=0.2198, simple_loss=0.3113, pruned_loss=0.06415, over 8636.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2881, pruned_loss=0.06258, over 1547801.60 frames. ], batch size: 34, lr: 3.83e-03, grad_scale: 8.0 +2023-02-07 00:05:06,548 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154239.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:05:08,022 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7665, 1.9825, 1.7211, 2.2721, 1.0126, 1.4598, 1.6888, 1.9418], + device='cuda:0'), covar=tensor([0.0772, 0.0761, 0.0915, 0.0403, 0.1026, 0.1346, 0.0704, 0.0665], + device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0196, 0.0247, 0.0209, 0.0204, 0.0248, 0.0249, 0.0209], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 00:05:34,061 INFO [train.py:901] (0/4) Epoch 20, batch 700, loss[loss=0.2015, simple_loss=0.2754, pruned_loss=0.06378, over 7702.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2886, pruned_loss=0.06272, over 1561619.17 frames. ], batch size: 18, lr: 3.83e-03, grad_scale: 8.0 +2023-02-07 00:05:37,459 INFO [optim.py:369] (0/4) 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 00:05:38,980 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0434, 2.3949, 1.8126, 2.8456, 1.4355, 1.6614, 2.0417, 2.3804], + device='cuda:0'), covar=tensor([0.0688, 0.0675, 0.0865, 0.0340, 0.1025, 0.1259, 0.0802, 0.0704], + device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0196, 0.0247, 0.0209, 0.0205, 0.0248, 0.0249, 0.0209], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 00:05:40,210 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=154285.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:05:54,018 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154304.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:05:57,527 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154309.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:05:59,511 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5152, 1.5150, 1.8699, 1.3354, 1.2406, 1.8268, 0.1879, 1.1844], + device='cuda:0'), covar=tensor([0.1921, 0.1282, 0.0409, 0.0925, 0.2728, 0.0459, 0.2180, 0.1250], + device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0195, 0.0124, 0.0222, 0.0270, 0.0135, 0.0170, 0.0187], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 00:06:08,842 INFO [train.py:901] (0/4) Epoch 20, batch 750, loss[loss=0.2002, simple_loss=0.2692, pruned_loss=0.06565, over 7261.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2902, pruned_loss=0.0637, over 1574353.02 frames. ], batch size: 16, lr: 3.83e-03, grad_scale: 8.0 +2023-02-07 00:06:11,930 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-07 00:06:28,505 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154355.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 00:06:33,652 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 20, batch 800, loss[loss=0.2084, simple_loss=0.2996, pruned_loss=0.05856, over 8339.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2904, pruned_loss=0.06394, over 1582809.60 frames. ], batch size: 26, lr: 3.83e-03, grad_scale: 8.0 +2023-02-07 00:06:43,012 WARNING [train.py:1067] (0/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-07 00:06:47,166 INFO [optim.py:369] (0/4) 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 00:07:01,172 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154400.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:07:08,310 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=154411.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 00:07:19,178 INFO [train.py:901] (0/4) Epoch 20, batch 850, loss[loss=0.2116, simple_loss=0.3039, pruned_loss=0.05968, over 8105.00 frames. ], tot_loss[loss=0.2086, simple_loss=0.2902, pruned_loss=0.06352, over 1594480.15 frames. ], batch size: 23, lr: 3.83e-03, grad_scale: 8.0 +2023-02-07 00:07:27,220 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=154438.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:07:32,649 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=154446.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:07:52,873 INFO [train.py:901] (0/4) Epoch 20, batch 900, loss[loss=0.2021, simple_loss=0.2874, pruned_loss=0.0584, over 8246.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2894, pruned_loss=0.0636, over 1599857.08 frames. ], batch size: 22, lr: 3.83e-03, grad_scale: 8.0 +2023-02-07 00:07:56,205 INFO [optim.py:369] (0/4) 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 00:07:57,767 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0140, 2.2996, 1.8502, 2.7121, 1.2382, 1.6464, 1.9062, 2.2436], + device='cuda:0'), covar=tensor([0.0700, 0.0679, 0.0915, 0.0392, 0.1151, 0.1313, 0.0858, 0.0733], + device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0195, 0.0246, 0.0209, 0.0205, 0.0248, 0.0249, 0.0208], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 00:07:59,808 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3326, 1.7397, 2.6804, 1.2157, 1.9424, 1.7336, 1.4941, 1.8711], + device='cuda:0'), covar=tensor([0.2144, 0.2577, 0.0913, 0.4893, 0.2031, 0.3439, 0.2474, 0.2376], + device='cuda:0'), in_proj_covar=tensor([0.0519, 0.0586, 0.0553, 0.0631, 0.0639, 0.0587, 0.0524, 0.0630], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 00:08:04,389 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154492.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:08:29,164 INFO [train.py:901] (0/4) Epoch 20, batch 950, loss[loss=0.1994, simple_loss=0.2906, pruned_loss=0.05409, over 8330.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2897, pruned_loss=0.06315, over 1605558.02 frames. ], batch size: 25, lr: 3.83e-03, grad_scale: 8.0 +2023-02-07 00:08:29,367 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154526.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 00:08:48,718 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154553.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:08:54,159 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154561.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:09:00,273 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9344, 1.4968, 3.2146, 1.3488, 2.2102, 3.5106, 3.6780, 3.0002], + device='cuda:0'), covar=tensor([0.1179, 0.1899, 0.0354, 0.2286, 0.1170, 0.0262, 0.0514, 0.0597], + device='cuda:0'), in_proj_covar=tensor([0.0288, 0.0318, 0.0286, 0.0310, 0.0302, 0.0259, 0.0403, 0.0303], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 00:09:01,512 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 20, batch 1000, loss[loss=0.2558, simple_loss=0.331, pruned_loss=0.09032, over 8194.00 frames. ], tot_loss[loss=0.209, simple_loss=0.291, pruned_loss=0.06352, over 1608062.59 frames. ], batch size: 23, lr: 3.83e-03, grad_scale: 8.0 +2023-02-07 00:09:07,493 INFO [optim.py:369] (0/4) 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 00:09:08,974 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=154583.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:09:35,154 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-07 00:09:38,959 INFO [train.py:901] (0/4) Epoch 20, batch 1050, loss[loss=0.1856, simple_loss=0.2861, pruned_loss=0.04259, over 8255.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2917, pruned_loss=0.06389, over 1610951.41 frames. ], batch size: 24, lr: 3.83e-03, grad_scale: 8.0 +2023-02-07 00:09:49,451 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-07 00:09:52,626 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.74 vs. limit=2.0 +2023-02-07 00:09:53,684 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154646.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:09:54,889 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=154648.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:09:59,113 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=154653.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:10:01,431 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 20, batch 1100, loss[loss=0.2023, simple_loss=0.2884, pruned_loss=0.05812, over 8532.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2908, pruned_loss=0.06382, over 1616637.04 frames. ], batch size: 28, lr: 3.83e-03, grad_scale: 8.0 +2023-02-07 00:10:18,094 INFO [optim.py:369] (0/4) 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 00:10:18,318 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154681.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:10:29,739 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=154699.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 00:10:45,302 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-07 00:10:48,864 INFO [train.py:901] (0/4) Epoch 20, batch 1150, loss[loss=0.1991, simple_loss=0.2859, pruned_loss=0.05616, over 8188.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2913, pruned_loss=0.06366, over 1620663.27 frames. ], batch size: 23, lr: 3.82e-03, grad_scale: 8.0 +2023-02-07 00:10:57,846 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154738.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:10:59,083 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-07 00:11:16,273 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154763.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:11:19,770 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154768.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:11:21,236 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7077, 2.0835, 3.2823, 1.5336, 2.4451, 2.0714, 1.7908, 2.4556], + device='cuda:0'), covar=tensor([0.1736, 0.2436, 0.0718, 0.4179, 0.1605, 0.3044, 0.2089, 0.2011], + device='cuda:0'), in_proj_covar=tensor([0.0521, 0.0586, 0.0553, 0.0631, 0.0642, 0.0589, 0.0524, 0.0630], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 00:11:25,033 INFO [train.py:901] (0/4) Epoch 20, batch 1200, loss[loss=0.2334, simple_loss=0.318, pruned_loss=0.07438, over 8541.00 frames. ], tot_loss[loss=0.2107, simple_loss=0.2922, pruned_loss=0.06457, over 1622212.63 frames. ], batch size: 39, lr: 3.82e-03, grad_scale: 8.0 +2023-02-07 00:11:28,383 INFO [optim.py:369] (0/4) 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 00:11:29,318 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154782.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 00:11:46,342 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154807.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 00:11:47,772 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154809.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:11:48,781 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.83 vs. limit=5.0 +2023-02-07 00:11:51,180 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154814.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 00:11:53,298 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 20, batch 1250, loss[loss=0.1593, simple_loss=0.2376, pruned_loss=0.04053, over 6807.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2916, pruned_loss=0.06463, over 1618094.09 frames. ], batch size: 15, lr: 3.82e-03, grad_scale: 8.0 +2023-02-07 00:12:05,340 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154834.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:12:06,466 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=154836.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:12:11,369 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154842.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:12:34,986 INFO [train.py:901] (0/4) Epoch 20, batch 1300, loss[loss=0.2141, simple_loss=0.3026, pruned_loss=0.06279, over 8289.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2921, pruned_loss=0.06441, over 1619964.76 frames. ], batch size: 23, lr: 3.82e-03, grad_scale: 8.0 +2023-02-07 00:12:38,324 INFO [optim.py:369] (0/4) 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 00:13:09,376 INFO [train.py:901] (0/4) Epoch 20, batch 1350, loss[loss=0.2329, simple_loss=0.3144, pruned_loss=0.07569, over 8557.00 frames. ], tot_loss[loss=0.2116, simple_loss=0.2931, pruned_loss=0.06504, over 1621313.07 frames. ], batch size: 49, lr: 3.82e-03, grad_scale: 8.0 +2023-02-07 00:13:27,094 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154951.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:13:29,046 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154954.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:13:44,745 INFO [train.py:901] (0/4) Epoch 20, batch 1400, loss[loss=0.1961, simple_loss=0.2791, pruned_loss=0.05657, over 8096.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2919, pruned_loss=0.06462, over 1616210.60 frames. ], batch size: 23, lr: 3.82e-03, grad_scale: 8.0 +2023-02-07 00:13:47,792 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154979.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:13:48,963 INFO [optim.py:369] (0/4) 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 00:13:55,291 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=154990.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:13:58,156 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154994.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:14:16,395 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155019.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:14:19,593 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155024.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:14:20,739 INFO [train.py:901] (0/4) Epoch 20, batch 1450, loss[loss=0.2138, simple_loss=0.3018, pruned_loss=0.06285, over 8254.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2931, pruned_loss=0.06542, over 1618245.23 frames. ], batch size: 24, lr: 3.82e-03, grad_scale: 8.0 +2023-02-07 00:14:29,164 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-07 00:14:33,294 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155044.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:14:36,494 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155049.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:14:50,888 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155070.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 00:14:55,240 INFO [train.py:901] (0/4) Epoch 20, batch 1500, loss[loss=0.2019, simple_loss=0.2855, pruned_loss=0.05912, over 7920.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2943, pruned_loss=0.06637, over 1619609.80 frames. ], batch size: 20, lr: 3.82e-03, grad_scale: 8.0 +2023-02-07 00:14:58,575 INFO [optim.py:369] (0/4) 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 00:14:59,295 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=155082.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:15:02,790 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155087.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:15:09,000 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155095.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 00:15:09,715 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.20 vs. limit=5.0 +2023-02-07 00:15:15,587 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155105.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:15:30,480 INFO [train.py:901] (0/4) Epoch 20, batch 1550, loss[loss=0.2476, simple_loss=0.3176, pruned_loss=0.08876, over 7020.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2936, pruned_loss=0.06521, over 1621279.68 frames. ], batch size: 71, lr: 3.82e-03, grad_scale: 8.0 +2023-02-07 00:15:36,919 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1281, 1.5307, 1.7518, 1.3972, 0.8846, 1.4671, 1.7547, 1.6482], + device='cuda:0'), covar=tensor([0.0529, 0.1242, 0.1586, 0.1414, 0.0634, 0.1470, 0.0682, 0.0618], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0151, 0.0189, 0.0157, 0.0100, 0.0160, 0.0111, 0.0141], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-07 00:16:04,725 INFO [train.py:901] (0/4) Epoch 20, batch 1600, loss[loss=0.1556, simple_loss=0.2358, pruned_loss=0.03768, over 6791.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2927, pruned_loss=0.0649, over 1622481.24 frames. ], batch size: 15, lr: 3.82e-03, grad_scale: 8.0 +2023-02-07 00:16:08,764 INFO [optim.py:369] (0/4) 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 00:16:20,605 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155197.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:16:27,181 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155207.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:16:40,713 INFO [train.py:901] (0/4) Epoch 20, batch 1650, loss[loss=0.1877, simple_loss=0.2724, pruned_loss=0.0515, over 8470.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2905, pruned_loss=0.06406, over 1619968.03 frames. ], batch size: 49, lr: 3.82e-03, grad_scale: 8.0 +2023-02-07 00:16:45,129 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155232.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:17:15,947 INFO [train.py:901] (0/4) Epoch 20, batch 1700, loss[loss=0.1754, simple_loss=0.2491, pruned_loss=0.0509, over 7713.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2902, pruned_loss=0.06379, over 1622867.31 frames. ], batch size: 18, lr: 3.82e-03, grad_scale: 8.0 +2023-02-07 00:17:19,372 INFO [optim.py:369] (0/4) 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 00:17:20,951 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4055, 1.7631, 1.8386, 1.1019, 1.8887, 1.3039, 0.4277, 1.6441], + device='cuda:0'), covar=tensor([0.0623, 0.0395, 0.0293, 0.0608, 0.0479, 0.0929, 0.0861, 0.0292], + device='cuda:0'), in_proj_covar=tensor([0.0443, 0.0384, 0.0338, 0.0442, 0.0370, 0.0529, 0.0388, 0.0409], + device='cuda:0'), out_proj_covar=tensor([1.1964e-04, 1.0103e-04, 8.9166e-05, 1.1711e-04, 9.7986e-05, 1.5067e-04, + 1.0512e-04, 1.0900e-04], device='cuda:0') +2023-02-07 00:17:44,035 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0210, 1.4765, 1.6146, 1.3046, 0.9668, 1.3534, 1.7860, 1.4497], + device='cuda:0'), covar=tensor([0.0514, 0.1250, 0.1691, 0.1456, 0.0596, 0.1544, 0.0678, 0.0665], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0151, 0.0190, 0.0158, 0.0100, 0.0161, 0.0112, 0.0141], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-07 00:17:51,286 INFO [train.py:901] (0/4) Epoch 20, batch 1750, loss[loss=0.2175, simple_loss=0.3076, pruned_loss=0.06372, over 8463.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2912, pruned_loss=0.06437, over 1621185.23 frames. ], batch size: 25, lr: 3.82e-03, grad_scale: 16.0 +2023-02-07 00:18:00,370 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=155338.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:18:11,541 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7215, 2.6748, 1.8118, 2.4149, 2.3449, 1.5688, 2.1496, 2.2741], + device='cuda:0'), covar=tensor([0.1453, 0.0408, 0.1270, 0.0639, 0.0642, 0.1533, 0.1014, 0.1001], + device='cuda:0'), in_proj_covar=tensor([0.0351, 0.0235, 0.0330, 0.0303, 0.0300, 0.0335, 0.0343, 0.0316], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 00:18:17,121 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155361.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:18:27,005 INFO [train.py:901] (0/4) Epoch 20, batch 1800, loss[loss=0.173, simple_loss=0.2694, pruned_loss=0.03831, over 7919.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2897, pruned_loss=0.06361, over 1618226.16 frames. ], batch size: 20, lr: 3.82e-03, grad_scale: 8.0 +2023-02-07 00:18:31,095 INFO [optim.py:369] (0/4) 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 00:18:34,029 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155386.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:19:01,127 INFO [train.py:901] (0/4) Epoch 20, batch 1850, loss[loss=0.2063, simple_loss=0.2958, pruned_loss=0.05838, over 8252.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2876, pruned_loss=0.06255, over 1614232.56 frames. ], batch size: 24, lr: 3.82e-03, grad_scale: 8.0 +2023-02-07 00:19:04,531 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=155431.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:19:20,143 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155453.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:19:20,187 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155453.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:19:36,697 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155475.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:19:37,256 INFO [train.py:901] (0/4) Epoch 20, batch 1900, loss[loss=0.2071, simple_loss=0.2851, pruned_loss=0.06454, over 8255.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2877, pruned_loss=0.06231, over 1615236.93 frames. ], batch size: 24, lr: 3.82e-03, grad_scale: 8.0 +2023-02-07 00:19:38,771 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155478.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:19:41,333 INFO [optim.py:369] (0/4) 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 00:20:11,843 INFO [train.py:901] (0/4) Epoch 20, batch 1950, loss[loss=0.18, simple_loss=0.277, pruned_loss=0.04152, over 8245.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2887, pruned_loss=0.06289, over 1615294.18 frames. ], batch size: 24, lr: 3.81e-03, grad_scale: 8.0 +2023-02-07 00:20:13,323 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-07 00:20:19,618 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7449, 1.5768, 1.7352, 1.4359, 0.8596, 1.5957, 1.6222, 1.4224], + device='cuda:0'), covar=tensor([0.0530, 0.1170, 0.1585, 0.1335, 0.0583, 0.1348, 0.0670, 0.0617], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0152, 0.0190, 0.0158, 0.0099, 0.0161, 0.0112, 0.0141], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-07 00:20:26,408 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155546.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:20:26,930 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-07 00:20:46,989 WARNING [train.py:1067] (0/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-07 00:20:47,681 INFO [train.py:901] (0/4) Epoch 20, batch 2000, loss[loss=0.2095, simple_loss=0.3038, pruned_loss=0.0576, over 8508.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2897, pruned_loss=0.06343, over 1615523.64 frames. ], batch size: 26, lr: 3.81e-03, grad_scale: 8.0 +2023-02-07 00:20:51,755 INFO [optim.py:369] (0/4) 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 00:21:20,964 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155623.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:21:22,881 INFO [train.py:901] (0/4) Epoch 20, batch 2050, loss[loss=0.2028, simple_loss=0.278, pruned_loss=0.06383, over 8361.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2893, pruned_loss=0.06338, over 1613353.10 frames. ], batch size: 24, lr: 3.81e-03, grad_scale: 8.0 +2023-02-07 00:21:40,403 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2922, 2.1207, 1.6088, 1.9675, 1.8263, 1.3708, 1.6254, 1.7208], + device='cuda:0'), covar=tensor([0.1250, 0.0425, 0.1322, 0.0552, 0.0646, 0.1561, 0.0916, 0.0944], + device='cuda:0'), in_proj_covar=tensor([0.0350, 0.0236, 0.0331, 0.0303, 0.0299, 0.0334, 0.0342, 0.0315], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 00:21:57,546 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155675.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:21:58,085 INFO [train.py:901] (0/4) Epoch 20, batch 2100, loss[loss=0.2276, simple_loss=0.3144, pruned_loss=0.07037, over 8348.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.29, pruned_loss=0.06373, over 1612059.19 frames. ], batch size: 24, lr: 3.81e-03, grad_scale: 8.0 +2023-02-07 00:22:02,096 INFO [optim.py:369] (0/4) 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 00:22:20,231 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-02-07 00:22:22,190 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155709.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:22:33,489 INFO [train.py:901] (0/4) Epoch 20, batch 2150, loss[loss=0.1993, simple_loss=0.2846, pruned_loss=0.05695, over 8290.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2904, pruned_loss=0.06399, over 1614428.36 frames. ], batch size: 23, lr: 3.81e-03, grad_scale: 8.0 +2023-02-07 00:22:39,018 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155734.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:23:07,311 INFO [train.py:901] (0/4) Epoch 20, batch 2200, loss[loss=0.2176, simple_loss=0.3054, pruned_loss=0.06489, over 8194.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2914, pruned_loss=0.06464, over 1612914.96 frames. ], batch size: 23, lr: 3.81e-03, grad_scale: 8.0 +2023-02-07 00:23:08,250 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9677, 1.6184, 3.4761, 1.6065, 2.3960, 3.8386, 3.9413, 3.3048], + device='cuda:0'), covar=tensor([0.1214, 0.1794, 0.0341, 0.2056, 0.1020, 0.0226, 0.0527, 0.0566], + device='cuda:0'), in_proj_covar=tensor([0.0289, 0.0319, 0.0285, 0.0311, 0.0301, 0.0261, 0.0407, 0.0302], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 00:23:12,093 INFO [optim.py:369] (0/4) 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 00:23:26,030 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155802.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:23:26,111 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155802.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:23:36,456 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1658, 1.6610, 4.4230, 1.9412, 2.5171, 4.9800, 5.0687, 4.3189], + device='cuda:0'), covar=tensor([0.1277, 0.1852, 0.0277, 0.2072, 0.1124, 0.0184, 0.0426, 0.0537], + device='cuda:0'), in_proj_covar=tensor([0.0289, 0.0320, 0.0285, 0.0312, 0.0301, 0.0261, 0.0407, 0.0302], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 00:23:38,410 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=155819.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:23:43,090 INFO [train.py:901] (0/4) Epoch 20, batch 2250, loss[loss=0.2102, simple_loss=0.3021, pruned_loss=0.05915, over 8498.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2916, pruned_loss=0.06502, over 1613356.31 frames. ], batch size: 26, lr: 3.81e-03, grad_scale: 8.0 +2023-02-07 00:23:44,827 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155827.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:24:17,857 INFO [train.py:901] (0/4) Epoch 20, batch 2300, loss[loss=0.2267, simple_loss=0.3006, pruned_loss=0.07636, over 7985.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2922, pruned_loss=0.06573, over 1609459.14 frames. ], batch size: 21, lr: 3.81e-03, grad_scale: 8.0 +2023-02-07 00:24:21,974 INFO [optim.py:369] (0/4) 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 00:24:54,606 INFO [train.py:901] (0/4) Epoch 20, batch 2350, loss[loss=0.1632, simple_loss=0.2496, pruned_loss=0.03836, over 7707.00 frames. ], tot_loss[loss=0.2107, simple_loss=0.2912, pruned_loss=0.06506, over 1608530.23 frames. ], batch size: 18, lr: 3.81e-03, grad_scale: 8.0 +2023-02-07 00:25:00,019 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155934.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:25:02,317 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.97 vs. limit=2.0 +2023-02-07 00:25:07,085 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([0.9418, 1.1948, 1.1976, 0.5830, 1.2243, 1.0089, 0.0562, 1.1463], + device='cuda:0'), covar=tensor([0.0500, 0.0388, 0.0356, 0.0589, 0.0406, 0.0976, 0.0865, 0.0307], + device='cuda:0'), in_proj_covar=tensor([0.0444, 0.0382, 0.0336, 0.0440, 0.0368, 0.0528, 0.0390, 0.0409], + device='cuda:0'), out_proj_covar=tensor([1.1971e-04, 1.0045e-04, 8.8696e-05, 1.1652e-04, 9.7482e-05, 1.5041e-04, + 1.0555e-04, 1.0907e-04], device='cuda:0') +2023-02-07 00:25:11,811 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4921, 1.7526, 2.5902, 1.4200, 1.9086, 1.8339, 1.5581, 1.9176], + device='cuda:0'), covar=tensor([0.1931, 0.2501, 0.0911, 0.4345, 0.1871, 0.3174, 0.2231, 0.2167], + device='cuda:0'), in_proj_covar=tensor([0.0515, 0.0585, 0.0550, 0.0628, 0.0637, 0.0586, 0.0521, 0.0625], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 00:25:23,244 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=155967.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:25:29,300 INFO [train.py:901] (0/4) Epoch 20, batch 2400, loss[loss=0.1761, simple_loss=0.2547, pruned_loss=0.04872, over 7684.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2915, pruned_loss=0.06463, over 1615421.03 frames. ], batch size: 18, lr: 3.81e-03, grad_scale: 8.0 +2023-02-07 00:25:33,216 INFO [optim.py:369] (0/4) 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 00:25:45,459 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-156000.pt +2023-02-07 00:26:00,990 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 20, batch 2450, loss[loss=0.1569, simple_loss=0.2382, pruned_loss=0.03778, over 7231.00 frames. ], tot_loss[loss=0.2107, simple_loss=0.2916, pruned_loss=0.06488, over 1614495.69 frames. ], batch size: 16, lr: 3.81e-03, grad_scale: 8.0 +2023-02-07 00:26:40,965 INFO [train.py:901] (0/4) Epoch 20, batch 2500, loss[loss=0.2205, simple_loss=0.3072, pruned_loss=0.06685, over 8545.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.292, pruned_loss=0.06508, over 1617749.16 frames. ], batch size: 49, lr: 3.81e-03, grad_scale: 8.0 +2023-02-07 00:26:45,023 INFO [optim.py:369] (0/4) 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 00:26:45,218 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156082.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:26:49,177 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156088.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:26:49,234 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3052, 3.7577, 2.4197, 3.1909, 3.1378, 2.1922, 2.9806, 3.3235], + device='cuda:0'), covar=tensor([0.1571, 0.0347, 0.1083, 0.0627, 0.0675, 0.1354, 0.1047, 0.0850], + device='cuda:0'), in_proj_covar=tensor([0.0350, 0.0235, 0.0329, 0.0305, 0.0297, 0.0334, 0.0340, 0.0315], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 00:27:15,806 INFO [train.py:901] (0/4) Epoch 20, batch 2550, loss[loss=0.2092, simple_loss=0.282, pruned_loss=0.06819, over 8077.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2909, pruned_loss=0.06453, over 1611760.01 frames. ], batch size: 21, lr: 3.81e-03, grad_scale: 8.0 +2023-02-07 00:27:21,370 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156134.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:27:29,834 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=156146.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:27:50,749 INFO [train.py:901] (0/4) Epoch 20, batch 2600, loss[loss=0.1921, simple_loss=0.2652, pruned_loss=0.05955, over 7292.00 frames. ], tot_loss[loss=0.2114, simple_loss=0.2918, pruned_loss=0.06548, over 1613581.54 frames. ], batch size: 16, lr: 3.81e-03, grad_scale: 8.0 +2023-02-07 00:27:54,662 INFO [optim.py:369] (0/4) 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 00:27:57,074 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.91 vs. limit=5.0 +2023-02-07 00:28:00,363 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156190.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:28:07,282 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1488, 1.9407, 2.5433, 2.1602, 2.4692, 2.2059, 1.9346, 1.4602], + device='cuda:0'), covar=tensor([0.5205, 0.4472, 0.1801, 0.3360, 0.2356, 0.2858, 0.2019, 0.4756], + device='cuda:0'), in_proj_covar=tensor([0.0941, 0.0971, 0.0797, 0.0935, 0.0995, 0.0883, 0.0745, 0.0824], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 00:28:12,720 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6944, 2.0763, 3.2761, 1.4770, 2.6522, 2.0891, 1.7808, 2.5396], + device='cuda:0'), covar=tensor([0.1858, 0.2549, 0.0942, 0.4346, 0.1718, 0.3103, 0.2111, 0.2169], + device='cuda:0'), in_proj_covar=tensor([0.0515, 0.0586, 0.0549, 0.0629, 0.0637, 0.0588, 0.0522, 0.0626], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 00:28:17,459 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156215.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:28:24,592 INFO [train.py:901] (0/4) Epoch 20, batch 2650, loss[loss=0.2067, simple_loss=0.2941, pruned_loss=0.05965, over 8464.00 frames. ], tot_loss[loss=0.2107, simple_loss=0.2914, pruned_loss=0.06494, over 1612692.75 frames. ], batch size: 27, lr: 3.81e-03, grad_scale: 8.0 +2023-02-07 00:28:30,659 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156234.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:28:49,175 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156261.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:29:00,116 INFO [train.py:901] (0/4) Epoch 20, batch 2700, loss[loss=0.2368, simple_loss=0.2969, pruned_loss=0.08834, over 7915.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2913, pruned_loss=0.0648, over 1618103.40 frames. ], batch size: 20, lr: 3.81e-03, grad_scale: 8.0 +2023-02-07 00:29:04,092 INFO [optim.py:369] (0/4) 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 00:29:23,238 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156308.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:29:35,128 INFO [train.py:901] (0/4) Epoch 20, batch 2750, loss[loss=0.2102, simple_loss=0.2771, pruned_loss=0.07163, over 6774.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2904, pruned_loss=0.06452, over 1610418.33 frames. ], batch size: 15, lr: 3.81e-03, grad_scale: 8.0 +2023-02-07 00:29:43,521 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156338.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:30:01,786 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156363.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:30:10,369 INFO [train.py:901] (0/4) Epoch 20, batch 2800, loss[loss=0.231, simple_loss=0.3136, pruned_loss=0.07417, over 8515.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2904, pruned_loss=0.06416, over 1608634.44 frames. ], batch size: 26, lr: 3.80e-03, grad_scale: 8.0 +2023-02-07 00:30:15,863 INFO [optim.py:369] (0/4) 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 00:30:20,866 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156390.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:30:38,812 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156415.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:30:42,967 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5602, 1.8693, 1.9661, 1.2190, 2.0272, 1.5276, 0.4027, 1.7931], + device='cuda:0'), covar=tensor([0.0465, 0.0320, 0.0266, 0.0481, 0.0363, 0.0845, 0.0721, 0.0220], + device='cuda:0'), in_proj_covar=tensor([0.0443, 0.0382, 0.0334, 0.0437, 0.0365, 0.0527, 0.0386, 0.0406], + device='cuda:0'), out_proj_covar=tensor([1.1936e-04, 1.0056e-04, 8.8201e-05, 1.1570e-04, 9.6665e-05, 1.5016e-04, + 1.0456e-04, 1.0826e-04], device='cuda:0') +2023-02-07 00:30:46,260 INFO [train.py:901] (0/4) Epoch 20, batch 2850, loss[loss=0.2195, simple_loss=0.3038, pruned_loss=0.06757, over 8328.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2905, pruned_loss=0.06402, over 1609078.93 frames. ], batch size: 25, lr: 3.80e-03, grad_scale: 8.0 +2023-02-07 00:30:50,397 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=156432.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:31:03,903 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.8574, 3.7969, 3.4619, 1.7792, 3.4180, 3.4845, 3.3838, 3.3387], + device='cuda:0'), covar=tensor([0.0827, 0.0673, 0.1144, 0.4601, 0.0920, 0.1015, 0.1462, 0.0877], + device='cuda:0'), in_proj_covar=tensor([0.0512, 0.0424, 0.0428, 0.0528, 0.0418, 0.0429, 0.0415, 0.0372], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 00:31:20,790 INFO [train.py:901] (0/4) Epoch 20, batch 2900, loss[loss=0.2225, simple_loss=0.308, pruned_loss=0.06848, over 8531.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2903, pruned_loss=0.06381, over 1610543.72 frames. ], batch size: 49, lr: 3.80e-03, grad_scale: 8.0 +2023-02-07 00:31:26,323 INFO [optim.py:369] (0/4) 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 00:31:50,393 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156517.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:31:53,709 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-07 00:31:57,135 INFO [train.py:901] (0/4) Epoch 20, batch 2950, loss[loss=0.1974, simple_loss=0.2861, pruned_loss=0.05433, over 8322.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2916, pruned_loss=0.06481, over 1610551.86 frames. ], batch size: 25, lr: 3.80e-03, grad_scale: 8.0 +2023-02-07 00:32:08,265 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156542.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:32:11,711 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2930, 1.9460, 4.3791, 2.0640, 2.3944, 4.9588, 5.0739, 4.1433], + device='cuda:0'), covar=tensor([0.1305, 0.1756, 0.0316, 0.2046, 0.1354, 0.0204, 0.0483, 0.0590], + device='cuda:0'), in_proj_covar=tensor([0.0292, 0.0323, 0.0287, 0.0315, 0.0305, 0.0262, 0.0410, 0.0305], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 00:32:11,735 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156547.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:32:31,025 INFO [train.py:901] (0/4) Epoch 20, batch 3000, loss[loss=0.2023, simple_loss=0.2902, pruned_loss=0.05719, over 8336.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2909, pruned_loss=0.06422, over 1613158.98 frames. ], batch size: 26, lr: 3.80e-03, grad_scale: 8.0 +2023-02-07 00:32:31,026 INFO [train.py:926] (0/4) Computing validation loss +2023-02-07 00:32:46,835 INFO [train.py:935] (0/4) Epoch 20, validation: loss=0.1756, simple_loss=0.2756, pruned_loss=0.03779, over 944034.00 frames. +2023-02-07 00:32:46,837 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-07 00:32:48,377 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=156578.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:32:51,797 INFO [optim.py:369] (0/4) 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 00:33:17,708 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.7554, 1.4542, 3.9240, 1.5438, 3.4458, 3.2512, 3.5810, 3.4222], + device='cuda:0'), covar=tensor([0.0720, 0.4370, 0.0644, 0.3810, 0.1215, 0.0978, 0.0644, 0.0780], + device='cuda:0'), in_proj_covar=tensor([0.0612, 0.0632, 0.0680, 0.0612, 0.0695, 0.0600, 0.0596, 0.0661], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 00:33:22,163 INFO [train.py:901] (0/4) Epoch 20, batch 3050, loss[loss=0.2878, simple_loss=0.36, pruned_loss=0.1078, over 8282.00 frames. ], tot_loss[loss=0.212, simple_loss=0.293, pruned_loss=0.06553, over 1617564.63 frames. ], batch size: 23, lr: 3.80e-03, grad_scale: 8.0 +2023-02-07 00:33:35,524 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.24 vs. limit=5.0 +2023-02-07 00:33:40,590 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=156652.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:33:40,697 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.8115, 1.5652, 1.9199, 1.5912, 0.9758, 1.6906, 2.0453, 2.0780], + device='cuda:0'), covar=tensor([0.0447, 0.1228, 0.1586, 0.1371, 0.0606, 0.1365, 0.0640, 0.0555], + device='cuda:0'), in_proj_covar=tensor([0.0098, 0.0153, 0.0192, 0.0159, 0.0101, 0.0162, 0.0112, 0.0142], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-07 00:33:57,464 INFO [train.py:901] (0/4) Epoch 20, batch 3100, loss[loss=0.2232, simple_loss=0.3065, pruned_loss=0.06998, over 8257.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2935, pruned_loss=0.06529, over 1620575.36 frames. ], batch size: 24, lr: 3.80e-03, grad_scale: 8.0 +2023-02-07 00:34:02,297 INFO [optim.py:369] (0/4) 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 00:34:09,231 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156693.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:34:31,976 INFO [train.py:901] (0/4) Epoch 20, batch 3150, loss[loss=0.2305, simple_loss=0.3119, pruned_loss=0.07454, over 8492.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2933, pruned_loss=0.06523, over 1620365.78 frames. ], batch size: 26, lr: 3.80e-03, grad_scale: 8.0 +2023-02-07 00:35:01,268 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156767.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:35:01,303 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156767.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:35:07,253 INFO [train.py:901] (0/4) Epoch 20, batch 3200, loss[loss=0.2642, simple_loss=0.3275, pruned_loss=0.1004, over 6915.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2914, pruned_loss=0.06432, over 1617730.11 frames. ], batch size: 72, lr: 3.80e-03, grad_scale: 8.0 +2023-02-07 00:35:11,890 INFO [optim.py:369] (0/4) 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 00:35:14,168 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5502, 1.5718, 2.0985, 1.3803, 1.1971, 2.0250, 0.3471, 1.2373], + device='cuda:0'), covar=tensor([0.1945, 0.1537, 0.0421, 0.1395, 0.2908, 0.0515, 0.2171, 0.1471], + device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0193, 0.0124, 0.0219, 0.0267, 0.0133, 0.0167, 0.0188], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 00:35:26,507 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156803.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:35:41,821 INFO [train.py:901] (0/4) Epoch 20, batch 3250, loss[loss=0.1922, simple_loss=0.2714, pruned_loss=0.05653, over 8131.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2902, pruned_loss=0.06377, over 1614376.13 frames. ], batch size: 22, lr: 3.80e-03, grad_scale: 8.0 +2023-02-07 00:35:43,296 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156828.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:36:17,054 INFO [train.py:901] (0/4) Epoch 20, batch 3300, loss[loss=0.202, simple_loss=0.2895, pruned_loss=0.05725, over 8360.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2903, pruned_loss=0.06372, over 1617221.22 frames. ], batch size: 26, lr: 3.80e-03, grad_scale: 8.0 +2023-02-07 00:36:21,768 INFO [optim.py:369] (0/4) 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 00:36:35,541 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156903.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:36:51,533 INFO [train.py:901] (0/4) Epoch 20, batch 3350, loss[loss=0.2013, simple_loss=0.2795, pruned_loss=0.06153, over 7928.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2905, pruned_loss=0.06352, over 1622331.69 frames. ], batch size: 20, lr: 3.80e-03, grad_scale: 8.0 +2023-02-07 00:37:07,172 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156949.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:37:11,494 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9587, 2.1556, 1.7172, 2.6635, 1.2194, 1.6158, 1.8590, 2.1991], + device='cuda:0'), covar=tensor([0.0753, 0.0789, 0.0978, 0.0409, 0.1161, 0.1339, 0.0902, 0.0752], + device='cuda:0'), in_proj_covar=tensor([0.0236, 0.0199, 0.0250, 0.0215, 0.0208, 0.0252, 0.0255, 0.0211], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 00:37:25,821 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156974.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:37:27,095 INFO [train.py:901] (0/4) Epoch 20, batch 3400, loss[loss=0.2325, simple_loss=0.3097, pruned_loss=0.0777, over 8240.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2915, pruned_loss=0.06414, over 1624182.10 frames. ], batch size: 22, lr: 3.80e-03, grad_scale: 8.0 +2023-02-07 00:37:31,900 INFO [optim.py:369] (0/4) 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 00:37:49,719 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.8287, 1.6197, 1.8636, 1.6459, 0.9188, 1.5857, 2.1605, 2.0056], + device='cuda:0'), covar=tensor([0.0451, 0.1232, 0.1681, 0.1383, 0.0629, 0.1481, 0.0641, 0.0568], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0153, 0.0191, 0.0158, 0.0100, 0.0161, 0.0112, 0.0142], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-07 00:37:59,111 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6220, 1.3470, 1.5483, 1.2357, 0.9180, 1.3514, 1.5568, 1.2773], + device='cuda:0'), covar=tensor([0.0556, 0.1299, 0.1691, 0.1479, 0.0599, 0.1552, 0.0694, 0.0682], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0153, 0.0191, 0.0159, 0.0100, 0.0161, 0.0112, 0.0142], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-07 00:38:01,289 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157023.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:38:03,918 INFO [train.py:901] (0/4) Epoch 20, batch 3450, loss[loss=0.1988, simple_loss=0.2709, pruned_loss=0.06338, over 7532.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2917, pruned_loss=0.06465, over 1619527.10 frames. ], batch size: 18, lr: 3.80e-03, grad_scale: 8.0 +2023-02-07 00:38:05,349 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157028.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:38:11,407 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.7011, 2.2261, 3.3257, 1.8670, 1.6738, 3.1658, 0.9107, 2.0791], + device='cuda:0'), covar=tensor([0.1604, 0.1392, 0.0351, 0.2022, 0.3047, 0.0423, 0.2374, 0.1775], + device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0191, 0.0123, 0.0216, 0.0265, 0.0132, 0.0165, 0.0186], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 00:38:18,895 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157048.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:38:38,200 INFO [train.py:901] (0/4) Epoch 20, batch 3500, loss[loss=0.1988, simple_loss=0.2802, pruned_loss=0.05869, over 8460.00 frames. ], tot_loss[loss=0.2114, simple_loss=0.2925, pruned_loss=0.06515, over 1617429.55 frames. ], batch size: 25, lr: 3.80e-03, grad_scale: 8.0 +2023-02-07 00:38:43,591 INFO [optim.py:369] (0/4) 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 00:38:55,890 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 +2023-02-07 00:39:02,270 WARNING [train.py:1067] (0/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-07 00:39:03,068 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=157111.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:39:09,294 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6541, 1.8902, 2.0774, 1.2577, 2.2033, 1.4717, 0.6359, 1.8677], + device='cuda:0'), covar=tensor([0.0615, 0.0370, 0.0284, 0.0588, 0.0391, 0.0923, 0.0865, 0.0306], + device='cuda:0'), in_proj_covar=tensor([0.0446, 0.0388, 0.0340, 0.0440, 0.0371, 0.0535, 0.0392, 0.0413], + device='cuda:0'), out_proj_covar=tensor([1.2021e-04, 1.0204e-04, 8.9864e-05, 1.1636e-04, 9.8216e-05, 1.5260e-04, + 1.0595e-04, 1.1022e-04], device='cuda:0') +2023-02-07 00:39:13,059 INFO [train.py:901] (0/4) Epoch 20, batch 3550, loss[loss=0.1958, simple_loss=0.2779, pruned_loss=0.05688, over 7929.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2922, pruned_loss=0.06511, over 1616468.54 frames. ], batch size: 20, lr: 3.80e-03, grad_scale: 8.0 +2023-02-07 00:39:37,677 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157160.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:39:48,301 INFO [train.py:901] (0/4) Epoch 20, batch 3600, loss[loss=0.2278, simple_loss=0.3092, pruned_loss=0.07314, over 8502.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2911, pruned_loss=0.06442, over 1614027.29 frames. ], batch size: 26, lr: 3.79e-03, grad_scale: 8.0 +2023-02-07 00:39:53,036 INFO [optim.py:369] (0/4) 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 00:40:10,603 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 +2023-02-07 00:40:15,592 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6425, 1.4583, 1.5955, 1.3640, 0.8945, 1.3972, 1.4992, 1.4138], + device='cuda:0'), covar=tensor([0.0647, 0.1265, 0.1695, 0.1462, 0.0622, 0.1533, 0.0736, 0.0664], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0152, 0.0190, 0.0159, 0.0100, 0.0161, 0.0112, 0.0141], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-07 00:40:24,181 INFO [train.py:901] (0/4) Epoch 20, batch 3650, loss[loss=0.2836, simple_loss=0.343, pruned_loss=0.1121, over 6985.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2909, pruned_loss=0.06412, over 1614192.76 frames. ], batch size: 71, lr: 3.79e-03, grad_scale: 8.0 +2023-02-07 00:40:24,360 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=157247.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:40:42,200 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 +2023-02-07 00:40:58,618 INFO [train.py:901] (0/4) Epoch 20, batch 3700, loss[loss=0.2125, simple_loss=0.2931, pruned_loss=0.06592, over 8329.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2922, pruned_loss=0.06447, over 1617943.77 frames. ], batch size: 25, lr: 3.79e-03, grad_scale: 8.0 +2023-02-07 00:41:03,187 INFO [optim.py:369] (0/4) 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 00:41:05,249 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 20, batch 3750, loss[loss=0.1931, simple_loss=0.2861, pruned_loss=0.05007, over 8702.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2932, pruned_loss=0.0652, over 1619274.69 frames. ], batch size: 39, lr: 3.79e-03, grad_scale: 8.0 +2023-02-07 00:41:58,948 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157362.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:42:05,450 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=157372.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:42:06,812 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157374.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:42:07,963 INFO [train.py:901] (0/4) Epoch 20, batch 3800, loss[loss=0.2163, simple_loss=0.302, pruned_loss=0.06529, over 8286.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2914, pruned_loss=0.0644, over 1618016.42 frames. ], batch size: 23, lr: 3.79e-03, grad_scale: 8.0 +2023-02-07 00:42:12,515 INFO [optim.py:369] (0/4) 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 00:42:25,022 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157400.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:42:29,222 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.10 vs. limit=5.0 +2023-02-07 00:42:39,614 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157422.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 00:42:42,793 INFO [train.py:901] (0/4) Epoch 20, batch 3850, loss[loss=0.2462, simple_loss=0.3258, pruned_loss=0.08329, over 8516.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2911, pruned_loss=0.06467, over 1612711.10 frames. ], batch size: 26, lr: 3.79e-03, grad_scale: 8.0 +2023-02-07 00:43:05,839 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.5562, 4.5244, 4.1084, 2.4330, 4.0383, 4.3046, 4.0988, 4.0457], + device='cuda:0'), covar=tensor([0.0772, 0.0577, 0.1168, 0.4053, 0.0769, 0.0804, 0.1197, 0.0653], + device='cuda:0'), in_proj_covar=tensor([0.0511, 0.0421, 0.0426, 0.0526, 0.0415, 0.0428, 0.0413, 0.0371], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 00:43:09,743 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-07 00:43:17,659 INFO [train.py:901] (0/4) Epoch 20, batch 3900, loss[loss=0.1584, simple_loss=0.2342, pruned_loss=0.0413, over 7655.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.291, pruned_loss=0.0646, over 1612230.25 frames. ], batch size: 19, lr: 3.79e-03, grad_scale: 8.0 +2023-02-07 00:43:21,784 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157482.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:43:22,201 INFO [optim.py:369] (0/4) 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 00:43:24,973 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157487.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:43:37,134 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=157504.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:43:39,407 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157507.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:43:43,541 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.1435, 1.3290, 4.3328, 1.5680, 3.7913, 3.6203, 3.9255, 3.7780], + device='cuda:0'), covar=tensor([0.0611, 0.4683, 0.0544, 0.4137, 0.1157, 0.0946, 0.0618, 0.0682], + device='cuda:0'), in_proj_covar=tensor([0.0612, 0.0635, 0.0687, 0.0616, 0.0697, 0.0603, 0.0599, 0.0665], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 00:43:51,264 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157524.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 00:43:52,537 INFO [train.py:901] (0/4) Epoch 20, batch 3950, loss[loss=0.1834, simple_loss=0.27, pruned_loss=0.04841, over 7538.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2914, pruned_loss=0.0646, over 1614559.09 frames. ], batch size: 18, lr: 3.79e-03, grad_scale: 8.0 +2023-02-07 00:44:25,099 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.2047, 4.1368, 3.8274, 1.9451, 3.7516, 3.8536, 3.7270, 3.6701], + device='cuda:0'), covar=tensor([0.0733, 0.0525, 0.1046, 0.4664, 0.0786, 0.0955, 0.1310, 0.0758], + device='cuda:0'), in_proj_covar=tensor([0.0516, 0.0425, 0.0431, 0.0531, 0.0418, 0.0432, 0.0418, 0.0374], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 00:44:28,111 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.79 vs. limit=5.0 +2023-02-07 00:44:28,450 INFO [train.py:901] (0/4) Epoch 20, batch 4000, loss[loss=0.1978, simple_loss=0.2662, pruned_loss=0.06474, over 7787.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2902, pruned_loss=0.06362, over 1613062.92 frames. ], batch size: 19, lr: 3.79e-03, grad_scale: 8.0 +2023-02-07 00:44:33,891 INFO [optim.py:369] (0/4) 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 00:44:34,064 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.6549, 1.7405, 4.8292, 1.7916, 4.3094, 4.0489, 4.4291, 4.3004], + device='cuda:0'), covar=tensor([0.0534, 0.4457, 0.0504, 0.4066, 0.0991, 0.0993, 0.0550, 0.0586], + device='cuda:0'), in_proj_covar=tensor([0.0614, 0.0637, 0.0688, 0.0619, 0.0701, 0.0606, 0.0601, 0.0667], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 00:44:36,139 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157586.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:44:41,348 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6379, 2.6939, 1.9498, 2.3901, 2.2195, 1.6882, 2.2889, 2.3237], + device='cuda:0'), covar=tensor([0.1352, 0.0341, 0.1045, 0.0538, 0.0706, 0.1397, 0.0865, 0.0849], + device='cuda:0'), in_proj_covar=tensor([0.0353, 0.0235, 0.0332, 0.0307, 0.0300, 0.0335, 0.0344, 0.0318], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 00:44:57,783 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157618.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:44:58,431 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157619.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:45:03,486 INFO [train.py:901] (0/4) Epoch 20, batch 4050, loss[loss=0.1794, simple_loss=0.2595, pruned_loss=0.04961, over 7795.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2906, pruned_loss=0.06362, over 1614143.72 frames. ], batch size: 19, lr: 3.79e-03, grad_scale: 8.0 +2023-02-07 00:45:15,102 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157643.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:45:29,208 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 +2023-02-07 00:45:33,138 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.83 vs. limit=5.0 +2023-02-07 00:45:38,035 INFO [train.py:901] (0/4) Epoch 20, batch 4100, loss[loss=0.2366, simple_loss=0.3227, pruned_loss=0.07523, over 8368.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2905, pruned_loss=0.06355, over 1617117.60 frames. ], batch size: 24, lr: 3.79e-03, grad_scale: 8.0 +2023-02-07 00:45:42,595 INFO [optim.py:369] (0/4) 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 00:46:07,480 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=157718.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:46:12,788 INFO [train.py:901] (0/4) Epoch 20, batch 4150, loss[loss=0.2081, simple_loss=0.2892, pruned_loss=0.0635, over 8540.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2908, pruned_loss=0.06357, over 1618293.07 frames. ], batch size: 28, lr: 3.79e-03, grad_scale: 8.0 +2023-02-07 00:46:25,105 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157743.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:46:25,619 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=157744.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:46:40,577 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=157766.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 00:46:41,942 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157768.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:46:45,980 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3951, 2.0619, 2.7701, 2.3101, 2.7366, 2.4000, 2.0995, 1.5934], + device='cuda:0'), covar=tensor([0.5071, 0.4821, 0.1836, 0.3389, 0.2342, 0.2922, 0.1945, 0.5037], + device='cuda:0'), in_proj_covar=tensor([0.0934, 0.0966, 0.0788, 0.0931, 0.0984, 0.0882, 0.0737, 0.0815], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 00:46:46,937 INFO [train.py:901] (0/4) Epoch 20, batch 4200, loss[loss=0.2358, simple_loss=0.3245, pruned_loss=0.07351, over 8191.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2918, pruned_loss=0.06403, over 1622047.12 frames. ], batch size: 23, lr: 3.79e-03, grad_scale: 8.0 +2023-02-07 00:46:52,343 INFO [optim.py:369] (0/4) 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 00:47:08,779 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-07 00:47:10,326 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.6086, 1.4497, 4.7465, 1.8503, 4.2422, 3.9236, 4.3399, 4.1756], + device='cuda:0'), covar=tensor([0.0472, 0.4582, 0.0486, 0.3827, 0.0987, 0.0925, 0.0513, 0.0626], + device='cuda:0'), in_proj_covar=tensor([0.0619, 0.0640, 0.0694, 0.0622, 0.0703, 0.0610, 0.0605, 0.0670], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 00:47:23,338 INFO [train.py:901] (0/4) Epoch 20, batch 4250, loss[loss=0.1999, simple_loss=0.2847, pruned_loss=0.05754, over 7923.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2921, pruned_loss=0.06431, over 1618726.33 frames. ], batch size: 20, lr: 3.79e-03, grad_scale: 8.0 +2023-02-07 00:47:26,220 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.8806, 1.7141, 1.8981, 1.6330, 1.0125, 1.6599, 2.1112, 1.8975], + device='cuda:0'), covar=tensor([0.0435, 0.1290, 0.1669, 0.1374, 0.0594, 0.1450, 0.0632, 0.0643], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0153, 0.0191, 0.0159, 0.0100, 0.0162, 0.0113, 0.0142], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-07 00:47:28,307 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157833.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:47:32,318 WARNING [train.py:1067] (0/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-07 00:47:46,384 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157859.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:47:53,340 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=157868.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 00:47:55,479 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157871.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:47:58,152 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157875.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:47:58,656 INFO [train.py:901] (0/4) Epoch 20, batch 4300, loss[loss=0.1718, simple_loss=0.2585, pruned_loss=0.04251, over 8084.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2917, pruned_loss=0.06373, over 1617507.43 frames. ], batch size: 21, lr: 3.79e-03, grad_scale: 8.0 +2023-02-07 00:48:02,027 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157881.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 00:48:03,192 INFO [optim.py:369] (0/4) 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 00:48:15,333 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157900.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:48:17,322 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6480, 1.4473, 2.9244, 1.3351, 2.1144, 3.0385, 3.2288, 2.6322], + device='cuda:0'), covar=tensor([0.1135, 0.1571, 0.0352, 0.2100, 0.0905, 0.0301, 0.0632, 0.0561], + device='cuda:0'), in_proj_covar=tensor([0.0292, 0.0318, 0.0285, 0.0311, 0.0300, 0.0262, 0.0407, 0.0303], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 00:48:24,974 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 +2023-02-07 00:48:33,497 INFO [train.py:901] (0/4) Epoch 20, batch 4350, loss[loss=0.2857, simple_loss=0.3533, pruned_loss=0.1091, over 7025.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2928, pruned_loss=0.06472, over 1617966.83 frames. ], batch size: 72, lr: 3.79e-03, grad_scale: 8.0 +2023-02-07 00:48:36,277 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=157930.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:49:04,095 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 20, batch 4400, loss[loss=0.1691, simple_loss=0.2513, pruned_loss=0.04347, over 8081.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.292, pruned_loss=0.06407, over 1617504.14 frames. ], batch size: 21, lr: 3.79e-03, grad_scale: 8.0 +2023-02-07 00:49:13,795 INFO [optim.py:369] (0/4) 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 00:49:13,987 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157983.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 00:49:24,899 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-158000.pt +2023-02-07 00:49:44,328 INFO [train.py:901] (0/4) Epoch 20, batch 4450, loss[loss=0.2388, simple_loss=0.3168, pruned_loss=0.08036, over 8017.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2911, pruned_loss=0.06355, over 1617773.17 frames. ], batch size: 22, lr: 3.78e-03, grad_scale: 8.0 +2023-02-07 00:49:45,692 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-07 00:49:57,304 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158045.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:50:18,898 INFO [train.py:901] (0/4) Epoch 20, batch 4500, loss[loss=0.2294, simple_loss=0.298, pruned_loss=0.08044, over 7655.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2911, pruned_loss=0.06375, over 1614330.46 frames. ], batch size: 19, lr: 3.78e-03, grad_scale: 8.0 +2023-02-07 00:50:23,587 INFO [optim.py:369] (0/4) 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 00:50:27,969 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158089.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:50:39,142 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-07 00:50:45,149 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158114.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:50:45,854 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158115.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:50:53,561 INFO [train.py:901] (0/4) Epoch 20, batch 4550, loss[loss=0.1937, simple_loss=0.2797, pruned_loss=0.05384, over 8704.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.291, pruned_loss=0.0644, over 1612778.38 frames. ], batch size: 30, lr: 3.78e-03, grad_scale: 8.0 +2023-02-07 00:51:01,048 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158137.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 00:51:02,894 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=158140.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:51:02,943 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158140.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:51:18,407 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158162.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 00:51:28,240 INFO [train.py:901] (0/4) Epoch 20, batch 4600, loss[loss=0.2051, simple_loss=0.2932, pruned_loss=0.05847, over 8241.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2908, pruned_loss=0.06466, over 1612677.61 frames. ], batch size: 24, lr: 3.78e-03, grad_scale: 8.0 +2023-02-07 00:51:32,827 INFO [optim.py:369] (0/4) 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 00:51:54,922 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=158215.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:52:01,856 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7982, 1.7743, 2.3161, 1.6987, 1.3715, 2.3146, 0.7076, 1.6322], + device='cuda:0'), covar=tensor([0.1784, 0.1132, 0.0321, 0.1155, 0.2699, 0.0394, 0.2037, 0.1228], + device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0195, 0.0124, 0.0219, 0.0268, 0.0134, 0.0167, 0.0189], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 00:52:03,093 INFO [train.py:901] (0/4) Epoch 20, batch 4650, loss[loss=0.2187, simple_loss=0.2979, pruned_loss=0.06977, over 8657.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.29, pruned_loss=0.06444, over 1610664.81 frames. ], batch size: 34, lr: 3.78e-03, grad_scale: 8.0 +2023-02-07 00:52:12,047 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158239.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 00:52:30,111 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158264.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 00:52:37,867 INFO [train.py:901] (0/4) Epoch 20, batch 4700, loss[loss=0.1994, simple_loss=0.2902, pruned_loss=0.0543, over 8328.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2891, pruned_loss=0.06361, over 1607246.09 frames. ], batch size: 25, lr: 3.78e-03, grad_scale: 8.0 +2023-02-07 00:52:42,601 INFO [optim.py:369] (0/4) 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 00:52:55,809 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 20, batch 4750, loss[loss=0.2083, simple_loss=0.2826, pruned_loss=0.06705, over 8721.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2882, pruned_loss=0.06285, over 1609236.58 frames. ], batch size: 30, lr: 3.78e-03, grad_scale: 8.0 +2023-02-07 00:53:12,920 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158326.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:53:15,563 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158330.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:53:40,940 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-07 00:53:43,665 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-07 00:53:48,277 INFO [train.py:901] (0/4) Epoch 20, batch 4800, loss[loss=0.2583, simple_loss=0.3229, pruned_loss=0.09688, over 7021.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2878, pruned_loss=0.063, over 1604884.02 frames. ], batch size: 72, lr: 3.78e-03, grad_scale: 16.0 +2023-02-07 00:53:52,906 INFO [optim.py:369] (0/4) 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 00:54:06,928 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=158402.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:54:22,895 INFO [train.py:901] (0/4) Epoch 20, batch 4850, loss[loss=0.1892, simple_loss=0.2769, pruned_loss=0.05072, over 8627.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2886, pruned_loss=0.06333, over 1608836.09 frames. ], batch size: 39, lr: 3.78e-03, grad_scale: 16.0 +2023-02-07 00:54:23,837 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9900, 1.7227, 2.0278, 1.7973, 1.9829, 2.0458, 1.8369, 0.7894], + device='cuda:0'), covar=tensor([0.5367, 0.4480, 0.1913, 0.3554, 0.2548, 0.2964, 0.1957, 0.5132], + device='cuda:0'), in_proj_covar=tensor([0.0934, 0.0966, 0.0788, 0.0933, 0.0990, 0.0882, 0.0740, 0.0820], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 00:54:33,547 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-07 00:54:35,059 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9760, 1.7536, 3.4647, 1.5161, 2.3154, 3.7605, 3.9241, 3.1992], + device='cuda:0'), covar=tensor([0.1338, 0.1766, 0.0388, 0.2251, 0.1195, 0.0257, 0.0588, 0.0632], + device='cuda:0'), in_proj_covar=tensor([0.0293, 0.0320, 0.0287, 0.0313, 0.0303, 0.0263, 0.0410, 0.0305], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 00:54:51,891 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0894, 2.0130, 3.1573, 1.7187, 2.4241, 3.4748, 3.5108, 3.0088], + device='cuda:0'), covar=tensor([0.1140, 0.1439, 0.0453, 0.1951, 0.1085, 0.0241, 0.0595, 0.0541], + device='cuda:0'), in_proj_covar=tensor([0.0293, 0.0319, 0.0286, 0.0312, 0.0303, 0.0262, 0.0409, 0.0305], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 00:54:57,238 INFO [train.py:901] (0/4) Epoch 20, batch 4900, loss[loss=0.2445, simple_loss=0.3143, pruned_loss=0.08735, over 7079.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2891, pruned_loss=0.06383, over 1605921.12 frames. ], batch size: 71, lr: 3.78e-03, grad_scale: 16.0 +2023-02-07 00:55:02,466 INFO [optim.py:369] (0/4) 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 00:55:03,227 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=158484.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:55:32,884 INFO [train.py:901] (0/4) Epoch 20, batch 4950, loss[loss=0.1988, simple_loss=0.2841, pruned_loss=0.0567, over 7965.00 frames. ], tot_loss[loss=0.2086, simple_loss=0.2895, pruned_loss=0.06386, over 1607742.57 frames. ], batch size: 21, lr: 3.78e-03, grad_scale: 16.0 +2023-02-07 00:56:07,757 INFO [train.py:901] (0/4) Epoch 20, batch 5000, loss[loss=0.2214, simple_loss=0.2975, pruned_loss=0.07265, over 8476.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.29, pruned_loss=0.06384, over 1612006.48 frames. ], batch size: 25, lr: 3.78e-03, grad_scale: 16.0 +2023-02-07 00:56:12,218 INFO [optim.py:369] (0/4) 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 00:56:12,596 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.19 vs. limit=5.0 +2023-02-07 00:56:14,503 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158586.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:56:23,817 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158599.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:56:32,785 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158611.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:56:42,897 INFO [train.py:901] (0/4) Epoch 20, batch 5050, loss[loss=0.2869, simple_loss=0.349, pruned_loss=0.1124, over 8025.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2919, pruned_loss=0.06519, over 1612434.69 frames. ], batch size: 22, lr: 3.78e-03, grad_scale: 16.0 +2023-02-07 00:56:54,121 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7641, 1.7193, 2.3468, 1.4680, 1.3136, 2.2912, 0.5624, 1.3944], + device='cuda:0'), covar=tensor([0.2207, 0.1434, 0.0348, 0.1651, 0.2942, 0.0518, 0.2387, 0.1702], + device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0197, 0.0127, 0.0222, 0.0273, 0.0135, 0.0171, 0.0192], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 00:57:10,200 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-07 00:57:17,768 INFO [train.py:901] (0/4) Epoch 20, batch 5100, loss[loss=0.2523, simple_loss=0.3223, pruned_loss=0.09121, over 7147.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2905, pruned_loss=0.06451, over 1610165.89 frames. ], batch size: 71, lr: 3.78e-03, grad_scale: 16.0 +2023-02-07 00:57:23,330 INFO [optim.py:369] (0/4) 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 00:57:53,848 INFO [train.py:901] (0/4) Epoch 20, batch 5150, loss[loss=0.2184, simple_loss=0.2881, pruned_loss=0.07433, over 7976.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2903, pruned_loss=0.06427, over 1610306.86 frames. ], batch size: 21, lr: 3.78e-03, grad_scale: 16.0 +2023-02-07 00:58:07,543 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 20, batch 5200, loss[loss=0.2463, simple_loss=0.3165, pruned_loss=0.08807, over 8460.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2901, pruned_loss=0.06421, over 1610089.22 frames. ], batch size: 29, lr: 3.78e-03, grad_scale: 16.0 +2023-02-07 00:58:30,714 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7354, 1.8372, 1.5983, 2.5114, 1.2258, 1.4033, 1.7636, 1.9512], + device='cuda:0'), covar=tensor([0.0938, 0.0892, 0.1187, 0.0542, 0.1201, 0.1593, 0.0973, 0.0888], + device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0195, 0.0245, 0.0212, 0.0203, 0.0246, 0.0249, 0.0207], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 00:58:33,214 INFO [optim.py:369] (0/4) 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 00:58:41,624 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=158795.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:58:46,650 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=158801.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:58:52,344 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-02-07 00:59:03,978 INFO [train.py:901] (0/4) Epoch 20, batch 5250, loss[loss=0.2201, simple_loss=0.3058, pruned_loss=0.06717, over 8502.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2904, pruned_loss=0.06443, over 1611192.54 frames. ], batch size: 26, lr: 3.78e-03, grad_scale: 16.0 +2023-02-07 00:59:11,286 WARNING [train.py:1067] (0/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-07 00:59:22,852 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=158853.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 00:59:24,330 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158855.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:59:28,231 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 20, batch 5300, loss[loss=0.186, simple_loss=0.2642, pruned_loss=0.05386, over 7425.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2903, pruned_loss=0.06447, over 1612622.06 frames. ], batch size: 17, lr: 3.77e-03, grad_scale: 16.0 +2023-02-07 00:59:41,413 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158880.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 00:59:43,356 INFO [optim.py:369] (0/4) 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 00:59:52,521 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.04 vs. limit=5.0 +2023-02-07 01:00:13,209 INFO [train.py:901] (0/4) Epoch 20, batch 5350, loss[loss=0.1791, simple_loss=0.267, pruned_loss=0.04559, over 7655.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.291, pruned_loss=0.06438, over 1618144.30 frames. ], batch size: 19, lr: 3.77e-03, grad_scale: 16.0 +2023-02-07 01:00:15,060 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 +2023-02-07 01:00:27,705 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 +2023-02-07 01:00:48,002 INFO [train.py:901] (0/4) Epoch 20, batch 5400, loss[loss=0.1927, simple_loss=0.2588, pruned_loss=0.06329, over 7224.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2904, pruned_loss=0.06463, over 1615284.08 frames. ], batch size: 16, lr: 3.77e-03, grad_scale: 16.0 +2023-02-07 01:00:52,646 INFO [optim.py:369] (0/4) 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 01:01:16,690 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-07 01:01:22,919 INFO [train.py:901] (0/4) Epoch 20, batch 5450, loss[loss=0.1988, simple_loss=0.2936, pruned_loss=0.05196, over 8295.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2909, pruned_loss=0.06419, over 1617827.26 frames. ], batch size: 23, lr: 3.77e-03, grad_scale: 16.0 +2023-02-07 01:01:42,891 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3302, 2.0808, 2.7845, 2.2209, 2.7176, 2.3321, 2.0863, 1.6546], + device='cuda:0'), covar=tensor([0.4935, 0.4409, 0.1823, 0.3304, 0.2147, 0.2651, 0.1717, 0.4794], + device='cuda:0'), in_proj_covar=tensor([0.0925, 0.0958, 0.0784, 0.0924, 0.0977, 0.0874, 0.0732, 0.0809], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 01:01:57,422 INFO [train.py:901] (0/4) Epoch 20, batch 5500, loss[loss=0.2298, simple_loss=0.3112, pruned_loss=0.07416, over 8243.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2911, pruned_loss=0.06425, over 1615563.80 frames. ], batch size: 24, lr: 3.77e-03, grad_scale: 16.0 +2023-02-07 01:02:00,098 WARNING [train.py:1067] (0/4) 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] (0/4) 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 01:02:03,906 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.46 vs. limit=2.0 +2023-02-07 01:02:05,769 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159087.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:02:21,366 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 +2023-02-07 01:02:27,304 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159117.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:02:29,231 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7227, 2.3592, 4.2670, 1.5646, 3.2935, 2.3830, 1.9023, 3.1127], + device='cuda:0'), covar=tensor([0.1817, 0.2632, 0.0701, 0.4336, 0.1539, 0.3030, 0.2122, 0.2302], + device='cuda:0'), in_proj_covar=tensor([0.0520, 0.0590, 0.0554, 0.0634, 0.0643, 0.0589, 0.0528, 0.0629], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 01:02:32,429 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7902, 1.6760, 3.1679, 1.4699, 2.2901, 3.4668, 3.5196, 2.9679], + device='cuda:0'), covar=tensor([0.1175, 0.1492, 0.0333, 0.2002, 0.0870, 0.0227, 0.0546, 0.0522], + device='cuda:0'), in_proj_covar=tensor([0.0294, 0.0319, 0.0286, 0.0314, 0.0304, 0.0262, 0.0410, 0.0305], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 01:02:32,980 INFO [train.py:901] (0/4) Epoch 20, batch 5550, loss[loss=0.2046, simple_loss=0.2791, pruned_loss=0.06505, over 7532.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2915, pruned_loss=0.06403, over 1619851.34 frames. ], batch size: 18, lr: 3.77e-03, grad_scale: 16.0 +2023-02-07 01:02:41,915 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=159139.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:02:43,056 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 +2023-02-07 01:02:44,150 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159142.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:02:45,997 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=159145.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:02:50,311 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 +2023-02-07 01:03:08,170 INFO [train.py:901] (0/4) Epoch 20, batch 5600, loss[loss=0.2646, simple_loss=0.3377, pruned_loss=0.09579, over 8531.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.291, pruned_loss=0.06376, over 1619181.21 frames. ], batch size: 31, lr: 3.77e-03, grad_scale: 16.0 +2023-02-07 01:03:09,014 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 01:03:20,709 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6539, 2.0573, 3.2679, 1.4761, 2.4946, 2.1944, 1.7848, 2.5263], + device='cuda:0'), covar=tensor([0.1856, 0.2635, 0.0900, 0.4398, 0.1801, 0.3035, 0.2177, 0.2168], + device='cuda:0'), in_proj_covar=tensor([0.0519, 0.0589, 0.0555, 0.0634, 0.0643, 0.0589, 0.0528, 0.0630], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 01:03:23,289 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=159197.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 01:03:44,008 INFO [train.py:901] (0/4) Epoch 20, batch 5650, loss[loss=0.1931, simple_loss=0.2759, pruned_loss=0.0552, over 8034.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2904, pruned_loss=0.06378, over 1614353.71 frames. ], batch size: 22, lr: 3.77e-03, grad_scale: 16.0 +2023-02-07 01:04:03,403 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159254.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:04:04,621 WARNING [train.py:1067] (0/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-07 01:04:07,496 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159260.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:04:19,056 INFO [train.py:901] (0/4) Epoch 20, batch 5700, loss[loss=0.189, simple_loss=0.2768, pruned_loss=0.05057, over 8505.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2912, pruned_loss=0.06405, over 1615282.57 frames. ], batch size: 28, lr: 3.77e-03, grad_scale: 8.0 +2023-02-07 01:04:25,333 INFO [optim.py:369] (0/4) 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 01:04:42,302 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159308.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:04:45,039 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159312.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 01:04:54,514 INFO [train.py:901] (0/4) Epoch 20, batch 5750, loss[loss=0.1838, simple_loss=0.2857, pruned_loss=0.04097, over 8193.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2907, pruned_loss=0.06402, over 1614881.55 frames. ], batch size: 23, lr: 3.77e-03, grad_scale: 8.0 +2023-02-07 01:04:59,220 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 +2023-02-07 01:05:09,318 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-07 01:05:23,185 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4781, 1.6918, 4.5652, 2.1807, 2.5827, 5.2778, 5.2895, 4.5888], + device='cuda:0'), covar=tensor([0.1062, 0.1665, 0.0230, 0.1748, 0.1070, 0.0141, 0.0412, 0.0441], + device='cuda:0'), in_proj_covar=tensor([0.0291, 0.0316, 0.0284, 0.0311, 0.0302, 0.0260, 0.0407, 0.0301], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 01:05:29,348 INFO [train.py:901] (0/4) Epoch 20, batch 5800, loss[loss=0.2014, simple_loss=0.2911, pruned_loss=0.05581, over 8560.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2899, pruned_loss=0.06342, over 1612171.29 frames. ], batch size: 39, lr: 3.77e-03, grad_scale: 8.0 +2023-02-07 01:05:35,563 INFO [optim.py:369] (0/4) 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 01:05:56,875 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0063, 1.6559, 3.1231, 1.4133, 2.1740, 3.3608, 3.5487, 2.8826], + device='cuda:0'), covar=tensor([0.1050, 0.1541, 0.0372, 0.2131, 0.1064, 0.0253, 0.0476, 0.0548], + device='cuda:0'), in_proj_covar=tensor([0.0291, 0.0315, 0.0283, 0.0310, 0.0302, 0.0259, 0.0405, 0.0300], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 01:06:04,881 INFO [train.py:901] (0/4) Epoch 20, batch 5850, loss[loss=0.2067, simple_loss=0.3069, pruned_loss=0.05327, over 8460.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2898, pruned_loss=0.06318, over 1614878.57 frames. ], batch size: 25, lr: 3.77e-03, grad_scale: 8.0 +2023-02-07 01:06:08,463 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=159431.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:06:39,995 INFO [train.py:901] (0/4) Epoch 20, batch 5900, loss[loss=0.2336, simple_loss=0.3034, pruned_loss=0.08189, over 7274.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.29, pruned_loss=0.06317, over 1613564.06 frames. ], batch size: 74, lr: 3.77e-03, grad_scale: 8.0 +2023-02-07 01:06:43,334 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 +2023-02-07 01:06:45,630 INFO [optim.py:369] (0/4) 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 01:07:04,115 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159510.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:07:08,807 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159516.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:07:12,080 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=159521.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:07:15,457 INFO [train.py:901] (0/4) Epoch 20, batch 5950, loss[loss=0.2139, simple_loss=0.3059, pruned_loss=0.06098, over 8487.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.29, pruned_loss=0.0634, over 1613928.31 frames. ], batch size: 28, lr: 3.77e-03, grad_scale: 8.0 +2023-02-07 01:07:21,759 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159535.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:07:26,375 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159541.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:07:29,658 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159546.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:07:45,590 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159568.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 01:07:50,335 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.4258, 4.3643, 3.9852, 2.0668, 3.8955, 3.9394, 3.9331, 3.7063], + device='cuda:0'), covar=tensor([0.0718, 0.0563, 0.1047, 0.4097, 0.0876, 0.0993, 0.1245, 0.0868], + device='cuda:0'), in_proj_covar=tensor([0.0512, 0.0424, 0.0428, 0.0526, 0.0415, 0.0427, 0.0413, 0.0372], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 01:07:50,944 INFO [train.py:901] (0/4) Epoch 20, batch 6000, loss[loss=0.2643, simple_loss=0.3385, pruned_loss=0.09508, over 8437.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2904, pruned_loss=0.06369, over 1613416.06 frames. ], batch size: 27, lr: 3.77e-03, grad_scale: 8.0 +2023-02-07 01:07:50,945 INFO [train.py:926] (0/4) Computing validation loss +2023-02-07 01:08:04,190 INFO [train.py:935] (0/4) Epoch 20, validation: loss=0.175, simple_loss=0.275, pruned_loss=0.03755, over 944034.00 frames. +2023-02-07 01:08:04,192 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-07 01:08:09,556 INFO [optim.py:369] (0/4) 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 01:08:15,914 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159593.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 01:08:38,927 INFO [train.py:901] (0/4) Epoch 20, batch 6050, loss[loss=0.2025, simple_loss=0.2864, pruned_loss=0.05933, over 8559.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2886, pruned_loss=0.06276, over 1614952.69 frames. ], batch size: 31, lr: 3.77e-03, grad_scale: 8.0 +2023-02-07 01:08:45,968 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159636.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:08:55,611 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159649.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:08:55,912 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 +2023-02-07 01:08:57,640 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=159652.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:09:14,781 INFO [train.py:901] (0/4) Epoch 20, batch 6100, loss[loss=0.2005, simple_loss=0.2945, pruned_loss=0.05328, over 8483.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2888, pruned_loss=0.06272, over 1618410.00 frames. ], batch size: 28, lr: 3.76e-03, grad_scale: 8.0 +2023-02-07 01:09:21,006 INFO [optim.py:369] (0/4) 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 01:09:41,572 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-07 01:09:49,995 INFO [train.py:901] (0/4) Epoch 20, batch 6150, loss[loss=0.1862, simple_loss=0.2538, pruned_loss=0.05934, over 7428.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2905, pruned_loss=0.06329, over 1622457.54 frames. ], batch size: 17, lr: 3.76e-03, grad_scale: 8.0 +2023-02-07 01:10:18,345 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159767.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:10:24,860 INFO [train.py:901] (0/4) Epoch 20, batch 6200, loss[loss=0.2439, simple_loss=0.3211, pruned_loss=0.08334, over 8193.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2901, pruned_loss=0.06348, over 1615521.31 frames. ], batch size: 23, lr: 3.76e-03, grad_scale: 8.0 +2023-02-07 01:10:30,204 INFO [optim.py:369] (0/4) 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 01:10:43,484 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159802.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:11:00,335 INFO [train.py:901] (0/4) Epoch 20, batch 6250, loss[loss=0.24, simple_loss=0.326, pruned_loss=0.07701, over 8461.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2895, pruned_loss=0.06279, over 1615877.99 frames. ], batch size: 27, lr: 3.76e-03, grad_scale: 8.0 +2023-02-07 01:11:01,222 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159827.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:11:07,404 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 +2023-02-07 01:11:32,491 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159873.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:11:34,372 INFO [train.py:901] (0/4) Epoch 20, batch 6300, loss[loss=0.2156, simple_loss=0.2968, pruned_loss=0.06722, over 8649.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2898, pruned_loss=0.06335, over 1617487.54 frames. ], batch size: 34, lr: 3.76e-03, grad_scale: 8.0 +2023-02-07 01:11:40,342 INFO [optim.py:369] (0/4) 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 01:11:45,868 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159892.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:11:58,776 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2056, 2.0008, 2.5775, 1.6332, 1.5952, 2.4982, 1.1671, 2.0314], + device='cuda:0'), covar=tensor([0.1651, 0.1290, 0.0464, 0.1455, 0.2510, 0.0433, 0.2168, 0.1451], + device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0196, 0.0127, 0.0222, 0.0272, 0.0134, 0.0170, 0.0190], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 01:12:03,496 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 20, batch 6350, loss[loss=0.2047, simple_loss=0.2805, pruned_loss=0.06447, over 8137.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.29, pruned_loss=0.06385, over 1617363.66 frames. ], batch size: 22, lr: 3.76e-03, grad_scale: 8.0 +2023-02-07 01:12:10,571 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 20, batch 6400, loss[loss=0.2156, simple_loss=0.2957, pruned_loss=0.06774, over 7823.00 frames. ], tot_loss[loss=0.2086, simple_loss=0.29, pruned_loss=0.06357, over 1619488.73 frames. ], batch size: 20, lr: 3.76e-03, grad_scale: 8.0 +2023-02-07 01:12:48,758 INFO [optim.py:369] (0/4) 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 01:12:55,746 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=159993.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:13:00,484 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-160000.pt +2023-02-07 01:13:05,827 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-07 01:13:16,794 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=160023.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:13:18,638 INFO [train.py:901] (0/4) Epoch 20, batch 6450, loss[loss=0.2311, simple_loss=0.3068, pruned_loss=0.07769, over 7916.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2907, pruned_loss=0.06408, over 1620108.68 frames. ], batch size: 20, lr: 3.76e-03, grad_scale: 8.0 +2023-02-07 01:13:34,516 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 20, batch 6500, loss[loss=0.2091, simple_loss=0.2861, pruned_loss=0.06612, over 8198.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2915, pruned_loss=0.0644, over 1623417.24 frames. ], batch size: 23, lr: 3.76e-03, grad_scale: 8.0 +2023-02-07 01:13:59,465 INFO [optim.py:369] (0/4) 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 01:14:16,476 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=160108.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:14:29,725 INFO [train.py:901] (0/4) Epoch 20, batch 6550, loss[loss=0.1876, simple_loss=0.2654, pruned_loss=0.05489, over 7967.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2916, pruned_loss=0.06464, over 1618803.80 frames. ], batch size: 21, lr: 3.76e-03, grad_scale: 8.0 +2023-02-07 01:14:53,090 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-07 01:15:05,556 INFO [train.py:901] (0/4) Epoch 20, batch 6600, loss[loss=0.1762, simple_loss=0.2604, pruned_loss=0.04598, over 7546.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2918, pruned_loss=0.06465, over 1619100.50 frames. ], batch size: 18, lr: 3.76e-03, grad_scale: 8.0 +2023-02-07 01:15:10,792 INFO [optim.py:369] (0/4) 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 01:15:12,123 WARNING [train.py:1067] (0/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-07 01:15:33,421 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=160217.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:15:39,336 INFO [train.py:901] (0/4) Epoch 20, batch 6650, loss[loss=0.2072, simple_loss=0.2783, pruned_loss=0.06802, over 7554.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2912, pruned_loss=0.06465, over 1619217.39 frames. ], batch size: 18, lr: 3.76e-03, grad_scale: 8.0 +2023-02-07 01:15:56,480 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5628, 1.4632, 2.3647, 1.3305, 2.2316, 2.5551, 2.7008, 2.1466], + device='cuda:0'), covar=tensor([0.1026, 0.1313, 0.0465, 0.1975, 0.0722, 0.0359, 0.0652, 0.0648], + device='cuda:0'), in_proj_covar=tensor([0.0296, 0.0321, 0.0287, 0.0315, 0.0306, 0.0263, 0.0412, 0.0305], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 01:16:12,500 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 20, batch 6700, loss[loss=0.1949, simple_loss=0.271, pruned_loss=0.05938, over 7709.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2907, pruned_loss=0.0644, over 1619482.26 frames. ], batch size: 18, lr: 3.76e-03, grad_scale: 8.0 +2023-02-07 01:16:20,502 INFO [optim.py:369] (0/4) 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 01:16:48,908 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=160325.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 01:16:49,348 INFO [train.py:901] (0/4) Epoch 20, batch 6750, loss[loss=0.29, simple_loss=0.3559, pruned_loss=0.1121, over 8762.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2918, pruned_loss=0.06496, over 1619650.64 frames. ], batch size: 30, lr: 3.76e-03, grad_scale: 8.0 +2023-02-07 01:16:53,598 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=160332.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:17:16,326 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 20, batch 6800, loss[loss=0.2017, simple_loss=0.2851, pruned_loss=0.05917, over 7922.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2926, pruned_loss=0.06537, over 1621252.12 frames. ], batch size: 20, lr: 3.76e-03, grad_scale: 8.0 +2023-02-07 01:17:28,099 WARNING [train.py:1067] (0/4) 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] (0/4) 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 01:17:30,379 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 +2023-02-07 01:17:31,580 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=160387.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:17:33,645 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=160389.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:17:57,296 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.5649, 1.5727, 4.7323, 1.8841, 4.2556, 3.9454, 4.3226, 4.2006], + device='cuda:0'), covar=tensor([0.0555, 0.4540, 0.0608, 0.4060, 0.1125, 0.1015, 0.0603, 0.0655], + device='cuda:0'), in_proj_covar=tensor([0.0612, 0.0629, 0.0681, 0.0613, 0.0694, 0.0595, 0.0596, 0.0667], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 01:17:59,211 INFO [train.py:901] (0/4) Epoch 20, batch 6850, loss[loss=0.1831, simple_loss=0.2744, pruned_loss=0.04587, over 8080.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2915, pruned_loss=0.06463, over 1623665.12 frames. ], batch size: 21, lr: 3.76e-03, grad_scale: 8.0 +2023-02-07 01:18:00,886 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 +2023-02-07 01:18:19,437 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-07 01:18:34,204 INFO [train.py:901] (0/4) Epoch 20, batch 6900, loss[loss=0.1877, simple_loss=0.2673, pruned_loss=0.05408, over 7921.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2905, pruned_loss=0.0642, over 1620250.68 frames. ], batch size: 20, lr: 3.76e-03, grad_scale: 8.0 +2023-02-07 01:18:39,564 INFO [optim.py:369] (0/4) 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 01:18:43,804 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.1634, 1.3876, 1.6831, 1.2537, 0.6869, 1.3980, 1.1621, 1.0422], + device='cuda:0'), covar=tensor([0.0600, 0.1232, 0.1654, 0.1478, 0.0564, 0.1514, 0.0721, 0.0727], + device='cuda:0'), in_proj_covar=tensor([0.0096, 0.0152, 0.0189, 0.0159, 0.0099, 0.0162, 0.0112, 0.0141], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-07 01:19:08,572 INFO [train.py:901] (0/4) Epoch 20, batch 6950, loss[loss=0.2248, simple_loss=0.3038, pruned_loss=0.07295, over 8243.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2893, pruned_loss=0.06333, over 1615602.68 frames. ], batch size: 22, lr: 3.76e-03, grad_scale: 8.0 +2023-02-07 01:19:15,013 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8020, 1.6641, 2.4538, 1.6845, 1.2099, 2.4550, 0.5960, 1.5139], + device='cuda:0'), covar=tensor([0.1639, 0.1461, 0.0386, 0.1380, 0.3120, 0.0356, 0.2349, 0.1425], + device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0195, 0.0126, 0.0222, 0.0271, 0.0134, 0.0169, 0.0189], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 01:19:30,209 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-07 01:19:31,066 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2803, 2.1672, 1.6165, 1.8475, 1.8073, 1.3847, 1.7060, 1.6657], + device='cuda:0'), covar=tensor([0.1285, 0.0404, 0.1378, 0.0588, 0.0727, 0.1627, 0.0907, 0.0878], + device='cuda:0'), in_proj_covar=tensor([0.0356, 0.0238, 0.0335, 0.0311, 0.0303, 0.0340, 0.0347, 0.0319], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 01:19:42,959 INFO [train.py:901] (0/4) Epoch 20, batch 7000, loss[loss=0.2149, simple_loss=0.2979, pruned_loss=0.06594, over 8294.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2879, pruned_loss=0.06261, over 1613729.30 frames. ], batch size: 23, lr: 3.75e-03, grad_scale: 8.0 +2023-02-07 01:19:48,348 INFO [optim.py:369] (0/4) 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 01:19:49,224 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.6850, 2.3619, 3.2465, 2.5153, 3.1532, 2.5954, 2.3963, 1.8413], + device='cuda:0'), covar=tensor([0.5020, 0.4925, 0.1955, 0.4089, 0.2646, 0.2849, 0.1799, 0.5856], + device='cuda:0'), in_proj_covar=tensor([0.0927, 0.0966, 0.0792, 0.0929, 0.0982, 0.0879, 0.0738, 0.0817], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 01:19:52,061 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=160588.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:20:09,528 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 20, batch 7050, loss[loss=0.2211, simple_loss=0.3112, pruned_loss=0.06548, over 8243.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2904, pruned_loss=0.06372, over 1617263.28 frames. ], batch size: 49, lr: 3.75e-03, grad_scale: 8.0 +2023-02-07 01:20:24,697 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 +2023-02-07 01:20:30,598 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=160643.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:20:48,805 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=160668.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:20:49,341 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=160669.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 01:20:53,998 INFO [train.py:901] (0/4) Epoch 20, batch 7100, loss[loss=0.21, simple_loss=0.2969, pruned_loss=0.06159, over 8322.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2903, pruned_loss=0.06364, over 1614070.42 frames. ], batch size: 25, lr: 3.75e-03, grad_scale: 8.0 +2023-02-07 01:20:59,625 INFO [optim.py:369] (0/4) 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 01:21:02,177 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-02-07 01:21:29,426 INFO [train.py:901] (0/4) Epoch 20, batch 7150, loss[loss=0.2126, simple_loss=0.2973, pruned_loss=0.06396, over 8329.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2898, pruned_loss=0.06378, over 1610516.24 frames. ], batch size: 26, lr: 3.75e-03, grad_scale: 8.0 +2023-02-07 01:21:30,767 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.8706, 6.1014, 5.1561, 2.4077, 5.3611, 5.6340, 5.4437, 5.4792], + device='cuda:0'), covar=tensor([0.0492, 0.0354, 0.1001, 0.4285, 0.0694, 0.0725, 0.1092, 0.0580], + device='cuda:0'), in_proj_covar=tensor([0.0520, 0.0432, 0.0435, 0.0536, 0.0424, 0.0440, 0.0423, 0.0381], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 01:21:49,560 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 +2023-02-07 01:22:04,613 INFO [train.py:901] (0/4) Epoch 20, batch 7200, loss[loss=0.1802, simple_loss=0.27, pruned_loss=0.04525, over 7825.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2893, pruned_loss=0.06333, over 1611516.42 frames. ], batch size: 20, lr: 3.75e-03, grad_scale: 8.0 +2023-02-07 01:22:09,775 INFO [optim.py:369] (0/4) 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 01:22:09,967 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=160784.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 01:22:18,952 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 +2023-02-07 01:22:22,738 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0580, 1.6024, 1.3117, 1.5184, 1.3003, 1.1311, 1.3693, 1.3131], + device='cuda:0'), covar=tensor([0.1120, 0.0487, 0.1403, 0.0548, 0.0823, 0.1719, 0.0830, 0.0805], + device='cuda:0'), in_proj_covar=tensor([0.0354, 0.0235, 0.0332, 0.0306, 0.0301, 0.0336, 0.0344, 0.0317], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 01:22:39,218 INFO [train.py:901] (0/4) Epoch 20, batch 7250, loss[loss=0.217, simple_loss=0.2969, pruned_loss=0.06855, over 8609.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2907, pruned_loss=0.06433, over 1611720.55 frames. ], batch size: 34, lr: 3.75e-03, grad_scale: 8.0 +2023-02-07 01:23:13,929 INFO [train.py:901] (0/4) Epoch 20, batch 7300, loss[loss=0.2301, simple_loss=0.3097, pruned_loss=0.0752, over 8631.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.291, pruned_loss=0.06396, over 1612565.22 frames. ], batch size: 31, lr: 3.75e-03, grad_scale: 8.0 +2023-02-07 01:23:19,316 INFO [optim.py:369] (0/4) 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 01:23:28,358 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5551, 1.8334, 1.9257, 1.2307, 1.9888, 1.5320, 0.4706, 1.8061], + device='cuda:0'), covar=tensor([0.0537, 0.0357, 0.0266, 0.0516, 0.0413, 0.0869, 0.0875, 0.0254], + device='cuda:0'), in_proj_covar=tensor([0.0450, 0.0384, 0.0339, 0.0440, 0.0369, 0.0534, 0.0392, 0.0412], + device='cuda:0'), out_proj_covar=tensor([1.2107e-04, 1.0064e-04, 8.9304e-05, 1.1617e-04, 9.7327e-05, 1.5183e-04, + 1.0604e-04, 1.0947e-04], device='cuda:0') +2023-02-07 01:23:44,246 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=160919.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:23:48,785 INFO [train.py:901] (0/4) Epoch 20, batch 7350, loss[loss=0.1915, simple_loss=0.2638, pruned_loss=0.0596, over 7714.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.29, pruned_loss=0.06313, over 1610222.41 frames. ], batch size: 18, lr: 3.75e-03, grad_scale: 8.0 +2023-02-07 01:23:51,530 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.2951, 1.3392, 3.4204, 1.0830, 3.0297, 2.9007, 3.1423, 3.0549], + device='cuda:0'), covar=tensor([0.0752, 0.3873, 0.0778, 0.3873, 0.1397, 0.1061, 0.0728, 0.0841], + device='cuda:0'), in_proj_covar=tensor([0.0612, 0.0630, 0.0679, 0.0615, 0.0696, 0.0594, 0.0595, 0.0665], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 01:24:16,155 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-07 01:24:20,000 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 +2023-02-07 01:24:24,317 INFO [train.py:901] (0/4) Epoch 20, batch 7400, loss[loss=0.179, simple_loss=0.2622, pruned_loss=0.04792, over 8029.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2891, pruned_loss=0.06299, over 1606285.06 frames. ], batch size: 22, lr: 3.75e-03, grad_scale: 8.0 +2023-02-07 01:24:29,908 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=160983.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:24:30,422 INFO [optim.py:369] (0/4) 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 01:24:37,302 WARNING [train.py:1067] (0/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-07 01:24:59,976 INFO [train.py:901] (0/4) Epoch 20, batch 7450, loss[loss=0.2306, simple_loss=0.3118, pruned_loss=0.07469, over 8462.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2892, pruned_loss=0.0629, over 1609860.59 frames. ], batch size: 25, lr: 3.75e-03, grad_scale: 8.0 +2023-02-07 01:25:10,065 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161040.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 01:25:16,122 WARNING [train.py:1067] (0/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-07 01:25:28,612 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161065.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 01:25:35,939 INFO [train.py:901] (0/4) Epoch 20, batch 7500, loss[loss=0.2076, simple_loss=0.2898, pruned_loss=0.06272, over 8476.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2888, pruned_loss=0.0632, over 1604224.88 frames. ], batch size: 28, lr: 3.75e-03, grad_scale: 8.0 +2023-02-07 01:25:41,424 INFO [optim.py:369] (0/4) 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 01:26:11,146 INFO [train.py:901] (0/4) Epoch 20, batch 7550, loss[loss=0.1926, simple_loss=0.262, pruned_loss=0.0616, over 7539.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2907, pruned_loss=0.06448, over 1605114.18 frames. ], batch size: 18, lr: 3.75e-03, grad_scale: 8.0 +2023-02-07 01:26:46,339 INFO [train.py:901] (0/4) Epoch 20, batch 7600, loss[loss=0.1954, simple_loss=0.2808, pruned_loss=0.05505, over 8503.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2908, pruned_loss=0.06425, over 1612238.56 frames. ], batch size: 39, lr: 3.75e-03, grad_scale: 8.0 +2023-02-07 01:26:51,651 INFO [optim.py:369] (0/4) 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 01:27:20,287 INFO [train.py:901] (0/4) Epoch 20, batch 7650, loss[loss=0.2021, simple_loss=0.2917, pruned_loss=0.05623, over 8475.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2903, pruned_loss=0.06449, over 1608762.78 frames. ], batch size: 25, lr: 3.75e-03, grad_scale: 8.0 +2023-02-07 01:27:25,753 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161234.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:27:36,274 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161249.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:27:45,578 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=161263.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:27:54,268 INFO [train.py:901] (0/4) Epoch 20, batch 7700, loss[loss=0.1665, simple_loss=0.2448, pruned_loss=0.04407, over 7549.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2904, pruned_loss=0.06464, over 1609545.08 frames. ], batch size: 18, lr: 3.75e-03, grad_scale: 16.0 +2023-02-07 01:27:59,447 INFO [optim.py:369] (0/4) 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 01:28:19,966 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 +2023-02-07 01:28:25,746 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-07 01:28:29,787 INFO [train.py:901] (0/4) Epoch 20, batch 7750, loss[loss=0.2002, simple_loss=0.2849, pruned_loss=0.05778, over 8783.00 frames. ], tot_loss[loss=0.2107, simple_loss=0.2913, pruned_loss=0.06508, over 1614102.30 frames. ], batch size: 40, lr: 3.75e-03, grad_scale: 16.0 +2023-02-07 01:28:30,570 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=161327.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:29:05,270 INFO [train.py:901] (0/4) Epoch 20, batch 7800, loss[loss=0.2119, simple_loss=0.289, pruned_loss=0.06736, over 8081.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2903, pruned_loss=0.06457, over 1616106.24 frames. ], batch size: 21, lr: 3.75e-03, grad_scale: 16.0 +2023-02-07 01:29:06,850 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=161378.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:29:10,618 INFO [optim.py:369] (0/4) 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 01:29:34,443 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 20, batch 7850, loss[loss=0.1941, simple_loss=0.2771, pruned_loss=0.0556, over 8346.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2915, pruned_loss=0.06477, over 1623275.53 frames. ], batch size: 24, lr: 3.74e-03, grad_scale: 16.0 +2023-02-07 01:29:49,624 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=161442.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:30:12,437 INFO [train.py:901] (0/4) Epoch 20, batch 7900, loss[loss=0.1739, simple_loss=0.2637, pruned_loss=0.04207, over 8089.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2921, pruned_loss=0.0651, over 1622950.67 frames. ], batch size: 21, lr: 3.74e-03, grad_scale: 8.0 +2023-02-07 01:30:13,747 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161478.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:30:18,876 INFO [optim.py:369] (0/4) 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 01:30:30,598 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8941, 2.2027, 1.6770, 2.8072, 1.4049, 1.5846, 2.0799, 2.2162], + device='cuda:0'), covar=tensor([0.0990, 0.0801, 0.1199, 0.0415, 0.1176, 0.1507, 0.0996, 0.0958], + device='cuda:0'), in_proj_covar=tensor([0.0235, 0.0197, 0.0247, 0.0213, 0.0205, 0.0247, 0.0252, 0.0209], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 01:30:45,508 INFO [train.py:901] (0/4) Epoch 20, batch 7950, loss[loss=0.1969, simple_loss=0.2753, pruned_loss=0.05924, over 7935.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2916, pruned_loss=0.06474, over 1619528.43 frames. ], batch size: 20, lr: 3.74e-03, grad_scale: 8.0 +2023-02-07 01:30:59,396 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1091, 1.8469, 2.3932, 2.0274, 2.2437, 2.1503, 1.8902, 1.0847], + device='cuda:0'), covar=tensor([0.5022, 0.4170, 0.1684, 0.3229, 0.2250, 0.2703, 0.1786, 0.4504], + device='cuda:0'), in_proj_covar=tensor([0.0936, 0.0969, 0.0796, 0.0932, 0.0989, 0.0882, 0.0742, 0.0818], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 01:31:18,035 INFO [train.py:901] (0/4) Epoch 20, batch 8000, loss[loss=0.1878, simple_loss=0.2753, pruned_loss=0.05017, over 7933.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2912, pruned_loss=0.06441, over 1621492.65 frames. ], batch size: 20, lr: 3.74e-03, grad_scale: 8.0 +2023-02-07 01:31:19,437 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=161578.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:31:23,852 INFO [optim.py:369] (0/4) 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 01:31:29,203 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.88 vs. limit=5.0 +2023-02-07 01:31:29,404 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=161593.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:31:51,325 INFO [train.py:901] (0/4) Epoch 20, batch 8050, loss[loss=0.1845, simple_loss=0.2674, pruned_loss=0.05082, over 7226.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2898, pruned_loss=0.06411, over 1609172.53 frames. ], batch size: 16, lr: 3.74e-03, grad_scale: 8.0 +2023-02-07 01:31:57,051 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161634.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:32:05,180 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.7362, 1.7947, 2.0004, 1.7650, 1.1979, 1.8247, 2.4840, 2.1598], + device='cuda:0'), covar=tensor([0.0580, 0.1482, 0.1970, 0.1653, 0.0801, 0.1727, 0.0697, 0.0624], + device='cuda:0'), in_proj_covar=tensor([0.0096, 0.0152, 0.0189, 0.0158, 0.0099, 0.0162, 0.0112, 0.0142], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-07 01:32:15,041 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/epoch-20.pt +2023-02-07 01:32:28,744 WARNING [train.py:1067] (0/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-07 01:32:32,190 INFO [train.py:901] (0/4) Epoch 21, batch 0, loss[loss=0.177, simple_loss=0.2558, pruned_loss=0.04913, over 7419.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2558, pruned_loss=0.04913, over 7419.00 frames. ], batch size: 17, lr: 3.65e-03, grad_scale: 8.0 +2023-02-07 01:32:32,191 INFO [train.py:926] (0/4) Computing validation loss +2023-02-07 01:32:44,210 INFO [train.py:935] (0/4) Epoch 21, validation: loss=0.1763, simple_loss=0.2762, pruned_loss=0.03818, over 944034.00 frames. +2023-02-07 01:32:44,211 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-07 01:32:44,451 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161659.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:32:49,224 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.1508, 4.1493, 3.6677, 1.9223, 3.6555, 3.7576, 3.7028, 3.6508], + device='cuda:0'), covar=tensor([0.0835, 0.0620, 0.1237, 0.4811, 0.1020, 0.1208, 0.1429, 0.0792], + device='cuda:0'), in_proj_covar=tensor([0.0511, 0.0429, 0.0427, 0.0526, 0.0419, 0.0430, 0.0413, 0.0375], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 01:32:59,362 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-07 01:33:02,224 INFO [optim.py:369] (0/4) 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 01:33:07,287 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8266, 1.9581, 2.1637, 1.4890, 2.2855, 1.6728, 0.6911, 1.9517], + device='cuda:0'), covar=tensor([0.0483, 0.0361, 0.0286, 0.0467, 0.0392, 0.0715, 0.0847, 0.0266], + device='cuda:0'), in_proj_covar=tensor([0.0451, 0.0384, 0.0336, 0.0437, 0.0368, 0.0531, 0.0391, 0.0412], + device='cuda:0'), out_proj_covar=tensor([1.2127e-04, 1.0075e-04, 8.8586e-05, 1.1548e-04, 9.7045e-05, 1.5098e-04, + 1.0579e-04, 1.0949e-04], device='cuda:0') +2023-02-07 01:33:07,960 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=161693.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:33:11,598 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161698.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:33:18,569 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 21, batch 50, loss[loss=0.1948, simple_loss=0.2782, pruned_loss=0.0557, over 8597.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2873, pruned_loss=0.06301, over 363305.31 frames. ], batch size: 39, lr: 3.65e-03, grad_scale: 8.0 +2023-02-07 01:33:29,248 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161723.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:33:32,459 WARNING [train.py:1067] (0/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-07 01:33:56,015 INFO [train.py:901] (0/4) Epoch 21, batch 100, loss[loss=0.2198, simple_loss=0.2856, pruned_loss=0.07698, over 7799.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2881, pruned_loss=0.06378, over 640398.77 frames. ], batch size: 20, lr: 3.65e-03, grad_scale: 8.0 +2023-02-07 01:33:57,243 WARNING [train.py:1067] (0/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-07 01:33:58,643 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=161763.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:34:14,161 INFO [optim.py:369] (0/4) 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 01:34:16,981 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1498, 1.8107, 3.4565, 1.4430, 2.4109, 3.8498, 3.9947, 3.2840], + device='cuda:0'), covar=tensor([0.1141, 0.1619, 0.0347, 0.2319, 0.1098, 0.0223, 0.0499, 0.0568], + device='cuda:0'), in_proj_covar=tensor([0.0294, 0.0320, 0.0286, 0.0315, 0.0306, 0.0261, 0.0410, 0.0304], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 01:34:17,016 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1845, 2.2927, 1.8386, 2.8948, 1.2947, 1.7748, 2.1881, 2.2931], + device='cuda:0'), covar=tensor([0.0689, 0.0786, 0.0903, 0.0349, 0.1172, 0.1256, 0.0862, 0.0809], + device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0196, 0.0244, 0.0210, 0.0203, 0.0244, 0.0248, 0.0206], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 01:34:30,762 INFO [train.py:901] (0/4) Epoch 21, batch 150, loss[loss=0.1518, simple_loss=0.2316, pruned_loss=0.03602, over 7192.00 frames. ], tot_loss[loss=0.2107, simple_loss=0.2911, pruned_loss=0.06514, over 859232.96 frames. ], batch size: 16, lr: 3.65e-03, grad_scale: 8.0 +2023-02-07 01:34:33,322 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.53 vs. limit=5.0 +2023-02-07 01:34:39,729 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=161822.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:34:47,239 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161833.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:35:06,331 INFO [train.py:901] (0/4) Epoch 21, batch 200, loss[loss=0.1657, simple_loss=0.2525, pruned_loss=0.03941, over 7981.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2906, pruned_loss=0.06437, over 1025408.21 frames. ], batch size: 21, lr: 3.65e-03, grad_scale: 8.0 +2023-02-07 01:35:15,833 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8927, 2.2807, 4.2669, 1.4844, 2.9040, 2.2941, 1.9071, 2.6774], + device='cuda:0'), covar=tensor([0.1755, 0.2502, 0.0797, 0.4569, 0.1835, 0.3248, 0.2145, 0.2650], + device='cuda:0'), in_proj_covar=tensor([0.0523, 0.0594, 0.0551, 0.0636, 0.0642, 0.0594, 0.0530, 0.0631], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 01:35:19,142 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=161878.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:35:23,716 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 21, batch 250, loss[loss=0.2184, simple_loss=0.2979, pruned_loss=0.06941, over 8464.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2913, pruned_loss=0.06367, over 1162237.28 frames. ], batch size: 27, lr: 3.65e-03, grad_scale: 8.0 +2023-02-07 01:35:47,944 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-07 01:35:55,197 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8981, 1.8127, 2.9295, 2.1766, 2.5785, 1.8915, 1.6530, 1.2403], + device='cuda:0'), covar=tensor([0.6912, 0.6071, 0.1867, 0.4197, 0.3032, 0.4297, 0.3028, 0.5823], + device='cuda:0'), in_proj_covar=tensor([0.0933, 0.0967, 0.0794, 0.0929, 0.0983, 0.0879, 0.0740, 0.0815], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 01:35:57,078 WARNING [train.py:1067] (0/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-07 01:36:00,661 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161949.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:36:15,257 INFO [train.py:901] (0/4) Epoch 21, batch 300, loss[loss=0.2751, simple_loss=0.3262, pruned_loss=0.112, over 7789.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2924, pruned_loss=0.06409, over 1265785.41 frames. ], batch size: 19, lr: 3.65e-03, grad_scale: 8.0 +2023-02-07 01:36:19,008 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161964.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:36:26,612 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161974.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:36:33,765 INFO [optim.py:369] (0/4) 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 01:36:36,643 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161989.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:36:44,138 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-162000.pt +2023-02-07 01:36:51,876 INFO [train.py:901] (0/4) Epoch 21, batch 350, loss[loss=0.265, simple_loss=0.3373, pruned_loss=0.09638, over 8497.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2899, pruned_loss=0.0633, over 1337573.50 frames. ], batch size: 26, lr: 3.65e-03, grad_scale: 8.0 +2023-02-07 01:37:14,170 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 +2023-02-07 01:37:25,797 INFO [train.py:901] (0/4) Epoch 21, batch 400, loss[loss=0.2358, simple_loss=0.3216, pruned_loss=0.07502, over 8461.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2908, pruned_loss=0.06355, over 1404658.83 frames. ], batch size: 25, lr: 3.65e-03, grad_scale: 8.0 +2023-02-07 01:37:29,049 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 +2023-02-07 01:37:44,477 INFO [optim.py:369] (0/4) 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 01:37:52,938 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=162095.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:38:02,161 INFO [train.py:901] (0/4) Epoch 21, batch 450, loss[loss=0.1799, simple_loss=0.2508, pruned_loss=0.05456, over 7786.00 frames. ], tot_loss[loss=0.208, simple_loss=0.29, pruned_loss=0.06297, over 1447776.20 frames. ], batch size: 19, lr: 3.64e-03, grad_scale: 8.0 +2023-02-07 01:38:20,138 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=162134.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:38:27,155 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6934, 2.2046, 4.1733, 1.4778, 2.9503, 2.1943, 1.7407, 3.0093], + device='cuda:0'), covar=tensor([0.1803, 0.2654, 0.0733, 0.4373, 0.1774, 0.3181, 0.2216, 0.2203], + device='cuda:0'), in_proj_covar=tensor([0.0521, 0.0593, 0.0550, 0.0633, 0.0640, 0.0592, 0.0529, 0.0631], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 01:38:31,233 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.1585, 1.3787, 4.2995, 1.6706, 3.8394, 3.5624, 3.9322, 3.8124], + device='cuda:0'), covar=tensor([0.0602, 0.4837, 0.0568, 0.4097, 0.1080, 0.1002, 0.0600, 0.0701], + device='cuda:0'), in_proj_covar=tensor([0.0615, 0.0632, 0.0683, 0.0614, 0.0697, 0.0597, 0.0598, 0.0664], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 01:38:37,320 INFO [train.py:901] (0/4) Epoch 21, batch 500, loss[loss=0.1652, simple_loss=0.2425, pruned_loss=0.04394, over 7786.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2877, pruned_loss=0.06194, over 1481076.53 frames. ], batch size: 19, lr: 3.64e-03, grad_scale: 8.0 +2023-02-07 01:38:37,537 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=162159.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:38:50,071 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=162177.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:38:55,592 INFO [optim.py:369] (0/4) 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 01:38:58,683 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3685, 1.5865, 2.0906, 1.2660, 1.5203, 1.6216, 1.4666, 1.5431], + device='cuda:0'), covar=tensor([0.2232, 0.3040, 0.1065, 0.5253, 0.2206, 0.3869, 0.2712, 0.2306], + device='cuda:0'), in_proj_covar=tensor([0.0520, 0.0593, 0.0550, 0.0632, 0.0640, 0.0592, 0.0530, 0.0632], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 01:39:01,456 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=162193.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:39:06,995 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0599, 1.7688, 3.5776, 1.6344, 2.5395, 3.9739, 4.0880, 3.3735], + device='cuda:0'), covar=tensor([0.1211, 0.1584, 0.0299, 0.2052, 0.0958, 0.0217, 0.0513, 0.0521], + device='cuda:0'), in_proj_covar=tensor([0.0292, 0.0319, 0.0286, 0.0312, 0.0306, 0.0261, 0.0409, 0.0303], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 01:39:13,737 INFO [train.py:901] (0/4) Epoch 21, batch 550, loss[loss=0.1959, simple_loss=0.2672, pruned_loss=0.06227, over 7270.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2881, pruned_loss=0.06203, over 1511725.59 frames. ], batch size: 16, lr: 3.64e-03, grad_scale: 8.0 +2023-02-07 01:39:20,121 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=162218.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:39:42,252 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=162249.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:39:48,770 INFO [train.py:901] (0/4) Epoch 21, batch 600, loss[loss=0.1895, simple_loss=0.2761, pruned_loss=0.05144, over 7928.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2891, pruned_loss=0.06247, over 1540074.30 frames. ], batch size: 20, lr: 3.64e-03, grad_scale: 8.0 +2023-02-07 01:40:02,425 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-07 01:40:06,586 INFO [optim.py:369] (0/4) 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 01:40:11,405 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 21, batch 650, loss[loss=0.1869, simple_loss=0.2714, pruned_loss=0.05118, over 8242.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2879, pruned_loss=0.06178, over 1560615.35 frames. ], batch size: 22, lr: 3.64e-03, grad_scale: 8.0 +2023-02-07 01:40:43,636 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7565, 2.4386, 3.8577, 1.5126, 2.7360, 1.9137, 2.0914, 2.2959], + device='cuda:0'), covar=tensor([0.2024, 0.2146, 0.1070, 0.4423, 0.1857, 0.3627, 0.2122, 0.3077], + device='cuda:0'), in_proj_covar=tensor([0.0520, 0.0592, 0.0548, 0.0631, 0.0640, 0.0590, 0.0529, 0.0630], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 01:40:49,740 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.9613, 1.9216, 2.1544, 1.9589, 1.0999, 1.8613, 2.2374, 2.2543], + device='cuda:0'), covar=tensor([0.0434, 0.1155, 0.1534, 0.1259, 0.0577, 0.1331, 0.0607, 0.0561], + device='cuda:0'), in_proj_covar=tensor([0.0096, 0.0151, 0.0188, 0.0157, 0.0098, 0.0161, 0.0111, 0.0141], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-07 01:40:59,237 INFO [train.py:901] (0/4) Epoch 21, batch 700, loss[loss=0.2457, simple_loss=0.3196, pruned_loss=0.08588, over 6943.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2887, pruned_loss=0.06273, over 1569213.78 frames. ], batch size: 71, lr: 3.64e-03, grad_scale: 8.0 +2023-02-07 01:41:17,761 INFO [optim.py:369] (0/4) 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 01:41:34,557 INFO [train.py:901] (0/4) Epoch 21, batch 750, loss[loss=0.1737, simple_loss=0.2573, pruned_loss=0.04508, over 8335.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2883, pruned_loss=0.06249, over 1578835.73 frames. ], batch size: 25, lr: 3.64e-03, grad_scale: 8.0 +2023-02-07 01:41:40,441 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 +2023-02-07 01:41:45,475 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-07 01:41:54,302 WARNING [train.py:1067] (0/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-07 01:41:56,361 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=162439.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:42:01,039 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 +2023-02-07 01:42:11,048 INFO [train.py:901] (0/4) Epoch 21, batch 800, loss[loss=0.1954, simple_loss=0.2721, pruned_loss=0.05937, over 7693.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2894, pruned_loss=0.06312, over 1588539.26 frames. ], batch size: 18, lr: 3.64e-03, grad_scale: 8.0 +2023-02-07 01:42:29,940 INFO [optim.py:369] (0/4) 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 01:42:47,168 INFO [train.py:901] (0/4) Epoch 21, batch 850, loss[loss=0.2138, simple_loss=0.2893, pruned_loss=0.06914, over 8135.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2897, pruned_loss=0.06336, over 1595142.00 frames. ], batch size: 22, lr: 3.64e-03, grad_scale: 8.0 +2023-02-07 01:43:01,466 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=162529.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:43:07,185 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4217, 2.2449, 2.9842, 2.4545, 2.9094, 2.3750, 2.2545, 1.9082], + device='cuda:0'), covar=tensor([0.4678, 0.4549, 0.1607, 0.3308, 0.2222, 0.2807, 0.1732, 0.4696], + device='cuda:0'), in_proj_covar=tensor([0.0936, 0.0967, 0.0791, 0.0929, 0.0983, 0.0878, 0.0738, 0.0815], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 01:43:16,271 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=162548.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:43:21,104 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162554.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:43:24,453 INFO [train.py:901] (0/4) Epoch 21, batch 900, loss[loss=0.2216, simple_loss=0.2971, pruned_loss=0.07305, over 8524.00 frames. ], tot_loss[loss=0.2086, simple_loss=0.2903, pruned_loss=0.06343, over 1601071.79 frames. ], batch size: 39, lr: 3.64e-03, grad_scale: 8.0 +2023-02-07 01:43:34,360 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=162573.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:43:42,638 INFO [optim.py:369] (0/4) 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] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=162593.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:44:00,578 INFO [train.py:901] (0/4) Epoch 21, batch 950, loss[loss=0.2233, simple_loss=0.3101, pruned_loss=0.06828, over 8196.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2903, pruned_loss=0.06359, over 1605598.33 frames. ], batch size: 23, lr: 3.64e-03, grad_scale: 8.0 +2023-02-07 01:44:14,212 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-07 01:44:35,848 INFO [train.py:901] (0/4) Epoch 21, batch 1000, loss[loss=0.221, simple_loss=0.3012, pruned_loss=0.07038, over 8445.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2898, pruned_loss=0.06323, over 1611241.04 frames. ], batch size: 27, lr: 3.64e-03, grad_scale: 8.0 +2023-02-07 01:44:40,150 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9601, 1.5942, 3.2661, 1.4781, 2.2857, 3.5925, 3.6988, 3.0520], + device='cuda:0'), covar=tensor([0.1170, 0.1711, 0.0371, 0.2166, 0.1166, 0.0241, 0.0539, 0.0552], + device='cuda:0'), in_proj_covar=tensor([0.0292, 0.0319, 0.0286, 0.0311, 0.0307, 0.0261, 0.0410, 0.0302], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 01:44:48,948 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-07 01:44:55,205 INFO [optim.py:369] (0/4) 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 01:45:01,388 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-07 01:45:11,681 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162708.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:45:12,206 INFO [train.py:901] (0/4) Epoch 21, batch 1050, loss[loss=0.1857, simple_loss=0.2802, pruned_loss=0.04556, over 8191.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2909, pruned_loss=0.06345, over 1615925.10 frames. ], batch size: 23, lr: 3.64e-03, grad_scale: 8.0 +2023-02-07 01:45:46,485 INFO [train.py:901] (0/4) Epoch 21, batch 1100, loss[loss=0.2058, simple_loss=0.2961, pruned_loss=0.05774, over 8293.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2902, pruned_loss=0.06328, over 1617550.53 frames. ], batch size: 23, lr: 3.64e-03, grad_scale: 8.0 +2023-02-07 01:46:06,016 INFO [optim.py:369] (0/4) 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 01:46:14,528 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-07 01:46:23,792 INFO [train.py:901] (0/4) Epoch 21, batch 1150, loss[loss=0.2169, simple_loss=0.3015, pruned_loss=0.06619, over 8246.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2893, pruned_loss=0.06266, over 1618264.61 frames. ], batch size: 24, lr: 3.64e-03, grad_scale: 8.0 +2023-02-07 01:46:24,671 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=162810.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:46:29,073 INFO [scaling.py:679] (0/4) 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] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 +2023-02-07 01:46:42,971 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=162835.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:46:50,055 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=162845.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:46:59,722 INFO [train.py:901] (0/4) Epoch 21, batch 1200, loss[loss=0.193, simple_loss=0.2857, pruned_loss=0.0501, over 8512.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2895, pruned_loss=0.0631, over 1615764.64 frames. ], batch size: 26, lr: 3.64e-03, grad_scale: 8.0 +2023-02-07 01:47:09,523 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=162873.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:47:17,449 INFO [optim.py:369] (0/4) 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 01:47:36,405 INFO [train.py:901] (0/4) Epoch 21, batch 1250, loss[loss=0.2181, simple_loss=0.3165, pruned_loss=0.05987, over 8474.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2894, pruned_loss=0.0632, over 1614908.26 frames. ], batch size: 25, lr: 3.64e-03, grad_scale: 8.0 +2023-02-07 01:48:11,284 INFO [train.py:901] (0/4) Epoch 21, batch 1300, loss[loss=0.2255, simple_loss=0.3082, pruned_loss=0.07138, over 8460.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2896, pruned_loss=0.06359, over 1616986.37 frames. ], batch size: 25, lr: 3.64e-03, grad_scale: 8.0 +2023-02-07 01:48:14,770 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=162964.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:48:27,750 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 +2023-02-07 01:48:28,481 INFO [optim.py:369] (0/4) 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 01:48:29,598 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.99 vs. limit=5.0 +2023-02-07 01:48:30,724 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162988.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:48:31,445 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=162989.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:48:44,803 INFO [train.py:901] (0/4) Epoch 21, batch 1350, loss[loss=0.1649, simple_loss=0.2436, pruned_loss=0.04305, over 7444.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2895, pruned_loss=0.06312, over 1620186.09 frames. ], batch size: 17, lr: 3.63e-03, grad_scale: 8.0 +2023-02-07 01:48:45,179 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 +2023-02-07 01:49:09,961 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=2.08 vs. limit=2.0 +2023-02-07 01:49:21,737 INFO [train.py:901] (0/4) Epoch 21, batch 1400, loss[loss=0.2138, simple_loss=0.2943, pruned_loss=0.0667, over 8459.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2895, pruned_loss=0.06372, over 1619478.55 frames. ], batch size: 25, lr: 3.63e-03, grad_scale: 8.0 +2023-02-07 01:49:34,632 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 +2023-02-07 01:49:38,380 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.22 vs. limit=5.0 +2023-02-07 01:49:39,401 INFO [optim.py:369] (0/4) 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 01:49:46,311 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-07 01:49:55,422 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=163108.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:49:55,932 INFO [train.py:901] (0/4) Epoch 21, batch 1450, loss[loss=0.2324, simple_loss=0.3182, pruned_loss=0.07332, over 8101.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2909, pruned_loss=0.06345, over 1620184.11 frames. ], batch size: 23, lr: 3.63e-03, grad_scale: 8.0 +2023-02-07 01:50:31,010 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.1930, 1.0707, 1.3027, 1.0111, 0.9162, 1.3347, 0.1034, 0.9118], + device='cuda:0'), covar=tensor([0.1607, 0.1354, 0.0466, 0.0838, 0.2815, 0.0533, 0.2101, 0.1223], + device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0192, 0.0125, 0.0219, 0.0269, 0.0132, 0.0167, 0.0188], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 01:50:32,151 INFO [train.py:901] (0/4) Epoch 21, batch 1500, loss[loss=0.2032, simple_loss=0.285, pruned_loss=0.06074, over 7977.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2903, pruned_loss=0.06304, over 1616481.73 frames. ], batch size: 21, lr: 3.63e-03, grad_scale: 8.0 +2023-02-07 01:50:35,020 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5363, 2.0111, 2.1706, 1.2217, 2.3254, 1.5039, 0.7303, 1.7454], + device='cuda:0'), covar=tensor([0.0691, 0.0393, 0.0271, 0.0689, 0.0373, 0.0898, 0.1002, 0.0402], + device='cuda:0'), in_proj_covar=tensor([0.0450, 0.0389, 0.0337, 0.0440, 0.0371, 0.0534, 0.0389, 0.0418], + device='cuda:0'), 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:0') +2023-02-07 01:50:50,507 INFO [optim.py:369] (0/4) 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 01:50:53,285 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=163189.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:51:05,627 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2277, 2.2681, 1.9982, 2.9665, 1.2895, 1.7528, 2.0226, 2.2305], + device='cuda:0'), covar=tensor([0.0646, 0.0785, 0.0869, 0.0337, 0.1151, 0.1231, 0.0939, 0.0802], + device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0196, 0.0245, 0.0213, 0.0205, 0.0247, 0.0250, 0.0208], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 01:51:06,787 INFO [train.py:901] (0/4) Epoch 21, batch 1550, loss[loss=0.2376, simple_loss=0.3174, pruned_loss=0.07891, over 8507.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2897, pruned_loss=0.06259, over 1622478.35 frames. ], batch size: 28, lr: 3.63e-03, grad_scale: 8.0 +2023-02-07 01:51:31,313 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=163244.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:51:42,280 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 21, batch 1600, loss[loss=0.2007, simple_loss=0.2836, pruned_loss=0.05889, over 8241.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2885, pruned_loss=0.06178, over 1622638.94 frames. ], batch size: 22, lr: 3.63e-03, grad_scale: 8.0 +2023-02-07 01:51:47,156 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6472, 1.3287, 1.6250, 1.2154, 0.8029, 1.4094, 1.4213, 1.4465], + device='cuda:0'), covar=tensor([0.0597, 0.1394, 0.1844, 0.1631, 0.0655, 0.1581, 0.0759, 0.0676], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0152, 0.0190, 0.0158, 0.0099, 0.0162, 0.0112, 0.0142], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-07 01:51:50,591 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=163269.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:52:00,873 INFO [optim.py:369] (0/4) 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 01:52:14,552 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=163304.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:52:17,721 INFO [train.py:901] (0/4) Epoch 21, batch 1650, loss[loss=0.1733, simple_loss=0.2431, pruned_loss=0.05172, over 7239.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2882, pruned_loss=0.06218, over 1619803.64 frames. ], batch size: 16, lr: 3.63e-03, grad_scale: 8.0 +2023-02-07 01:52:51,344 INFO [train.py:901] (0/4) Epoch 21, batch 1700, loss[loss=0.1788, simple_loss=0.2484, pruned_loss=0.05461, over 7688.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2892, pruned_loss=0.06252, over 1620149.30 frames. ], batch size: 18, lr: 3.63e-03, grad_scale: 8.0 +2023-02-07 01:53:09,964 INFO [optim.py:369] (0/4) 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 01:53:27,421 INFO [train.py:901] (0/4) Epoch 21, batch 1750, loss[loss=0.2108, simple_loss=0.2889, pruned_loss=0.06636, over 8249.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2892, pruned_loss=0.06278, over 1619286.99 frames. ], batch size: 24, lr: 3.63e-03, grad_scale: 8.0 +2023-02-07 01:53:51,710 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4411, 2.0654, 2.1097, 1.9636, 1.4526, 1.9924, 2.2809, 2.0738], + device='cuda:0'), covar=tensor([0.0480, 0.0891, 0.1296, 0.1156, 0.0586, 0.1122, 0.0617, 0.0511], + device='cuda:0'), in_proj_covar=tensor([0.0096, 0.0151, 0.0188, 0.0158, 0.0098, 0.0161, 0.0112, 0.0142], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-07 01:53:56,328 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=163452.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:54:01,060 INFO [train.py:901] (0/4) Epoch 21, batch 1800, loss[loss=0.193, simple_loss=0.2747, pruned_loss=0.05563, over 7964.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2894, pruned_loss=0.06283, over 1618759.90 frames. ], batch size: 21, lr: 3.63e-03, grad_scale: 16.0 +2023-02-07 01:54:18,722 INFO [optim.py:369] (0/4) 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 01:54:36,585 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 +2023-02-07 01:54:37,343 INFO [train.py:901] (0/4) Epoch 21, batch 1850, loss[loss=0.207, simple_loss=0.2846, pruned_loss=0.0647, over 8197.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2905, pruned_loss=0.06354, over 1619720.67 frames. ], batch size: 23, lr: 3.63e-03, grad_scale: 16.0 +2023-02-07 01:55:11,694 INFO [train.py:901] (0/4) Epoch 21, batch 1900, loss[loss=0.2306, simple_loss=0.3069, pruned_loss=0.07714, over 6711.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2899, pruned_loss=0.0632, over 1621232.22 frames. ], batch size: 71, lr: 3.63e-03, grad_scale: 16.0 +2023-02-07 01:55:12,595 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=163560.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:55:17,223 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=163567.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:55:26,427 WARNING [train.py:1067] (0/4) 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] (0/4) 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 01:55:29,214 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=163585.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:55:37,692 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-07 01:55:40,606 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=163602.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:55:43,574 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 +2023-02-07 01:55:45,148 INFO [train.py:901] (0/4) Epoch 21, batch 1950, loss[loss=0.1957, simple_loss=0.2859, pruned_loss=0.05276, over 8513.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2895, pruned_loss=0.0631, over 1621240.55 frames. ], batch size: 28, lr: 3.63e-03, grad_scale: 16.0 +2023-02-07 01:55:58,504 WARNING [train.py:1067] (0/4) 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] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 +2023-02-07 01:56:21,642 INFO [train.py:901] (0/4) Epoch 21, batch 2000, loss[loss=0.2489, simple_loss=0.3186, pruned_loss=0.08963, over 8586.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2893, pruned_loss=0.06326, over 1618995.04 frames. ], batch size: 34, lr: 3.63e-03, grad_scale: 16.0 +2023-02-07 01:56:29,872 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6433, 1.8943, 2.0215, 1.4422, 2.1465, 1.4989, 0.5110, 1.8845], + device='cuda:0'), covar=tensor([0.0498, 0.0321, 0.0262, 0.0499, 0.0336, 0.0810, 0.0805, 0.0259], + device='cuda:0'), in_proj_covar=tensor([0.0446, 0.0384, 0.0336, 0.0438, 0.0367, 0.0528, 0.0387, 0.0414], + device='cuda:0'), 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:0') +2023-02-07 01:56:39,056 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 21, batch 2050, loss[loss=0.219, simple_loss=0.298, pruned_loss=0.06998, over 8034.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2893, pruned_loss=0.06301, over 1620900.30 frames. ], batch size: 22, lr: 3.63e-03, grad_scale: 16.0 +2023-02-07 01:57:00,599 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=163717.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:57:05,691 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.8313, 1.3092, 3.9729, 1.3982, 3.5388, 3.3220, 3.5794, 3.4808], + device='cuda:0'), covar=tensor([0.0655, 0.4594, 0.0696, 0.4149, 0.1179, 0.1080, 0.0662, 0.0750], + device='cuda:0'), in_proj_covar=tensor([0.0624, 0.0640, 0.0685, 0.0622, 0.0701, 0.0603, 0.0602, 0.0670], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 01:57:30,299 INFO [train.py:901] (0/4) Epoch 21, batch 2100, loss[loss=0.2189, simple_loss=0.3005, pruned_loss=0.06859, over 8345.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2892, pruned_loss=0.06307, over 1621871.13 frames. ], batch size: 26, lr: 3.63e-03, grad_scale: 16.0 +2023-02-07 01:57:43,622 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 +2023-02-07 01:57:48,642 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 21, batch 2150, loss[loss=0.2004, simple_loss=0.2766, pruned_loss=0.06209, over 7778.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2889, pruned_loss=0.06297, over 1620893.17 frames. ], batch size: 19, lr: 3.63e-03, grad_scale: 16.0 +2023-02-07 01:58:14,671 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=163823.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:58:31,299 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=163848.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 01:58:39,334 INFO [train.py:901] (0/4) Epoch 21, batch 2200, loss[loss=0.214, simple_loss=0.2897, pruned_loss=0.0691, over 6428.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2883, pruned_loss=0.06291, over 1610822.93 frames. ], batch size: 14, lr: 3.63e-03, grad_scale: 16.0 +2023-02-07 01:58:58,241 INFO [optim.py:369] (0/4) 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 01:59:15,135 INFO [train.py:901] (0/4) Epoch 21, batch 2250, loss[loss=0.2319, simple_loss=0.3156, pruned_loss=0.07408, over 8609.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2884, pruned_loss=0.06298, over 1610751.62 frames. ], batch size: 34, lr: 3.62e-03, grad_scale: 16.0 +2023-02-07 01:59:21,982 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.8058, 3.7682, 3.3842, 2.0549, 3.3128, 3.4134, 3.3458, 3.2687], + device='cuda:0'), covar=tensor([0.0934, 0.0722, 0.1177, 0.4741, 0.1047, 0.1288, 0.1480, 0.1026], + device='cuda:0'), in_proj_covar=tensor([0.0519, 0.0433, 0.0435, 0.0538, 0.0426, 0.0441, 0.0421, 0.0382], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 01:59:30,190 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.7591, 5.9716, 5.0557, 3.0491, 5.1816, 5.5991, 5.4676, 5.3551], + device='cuda:0'), covar=tensor([0.0601, 0.0396, 0.0980, 0.3826, 0.0846, 0.0775, 0.1014, 0.0743], + device='cuda:0'), in_proj_covar=tensor([0.0520, 0.0433, 0.0435, 0.0538, 0.0427, 0.0441, 0.0421, 0.0383], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 01:59:33,084 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7624, 1.9775, 2.1138, 1.4773, 2.2729, 1.4981, 0.7053, 1.9508], + device='cuda:0'), covar=tensor([0.0599, 0.0318, 0.0274, 0.0550, 0.0359, 0.0858, 0.0904, 0.0290], + device='cuda:0'), in_proj_covar=tensor([0.0448, 0.0384, 0.0336, 0.0439, 0.0369, 0.0526, 0.0387, 0.0413], + device='cuda:0'), 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:0') +2023-02-07 01:59:49,152 INFO [train.py:901] (0/4) Epoch 21, batch 2300, loss[loss=0.1882, simple_loss=0.2758, pruned_loss=0.05033, over 8587.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2886, pruned_loss=0.06281, over 1609578.44 frames. ], batch size: 31, lr: 3.62e-03, grad_scale: 16.0 +2023-02-07 01:59:58,901 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=163998.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:00:19,119 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-164000.pt +2023-02-07 02:00:20,572 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 +2023-02-07 02:00:26,179 INFO [train.py:901] (0/4) Epoch 21, batch 2350, loss[loss=0.171, simple_loss=0.2488, pruned_loss=0.04656, over 7230.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2877, pruned_loss=0.06231, over 1610224.85 frames. ], batch size: 16, lr: 3.62e-03, grad_scale: 16.0 +2023-02-07 02:00:39,153 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 +2023-02-07 02:00:46,866 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0358, 1.2081, 1.1715, 0.8319, 1.1816, 1.0219, 0.1106, 1.1668], + device='cuda:0'), covar=tensor([0.0412, 0.0375, 0.0332, 0.0482, 0.0376, 0.0898, 0.0807, 0.0313], + device='cuda:0'), in_proj_covar=tensor([0.0447, 0.0383, 0.0334, 0.0437, 0.0368, 0.0525, 0.0386, 0.0412], + device='cuda:0'), 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:0') +2023-02-07 02:01:01,226 INFO [train.py:901] (0/4) Epoch 21, batch 2400, loss[loss=0.2132, simple_loss=0.2903, pruned_loss=0.06806, over 7792.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2883, pruned_loss=0.06283, over 1609322.45 frames. ], batch size: 19, lr: 3.62e-03, grad_scale: 16.0 +2023-02-07 02:01:19,280 INFO [optim.py:369] (0/4) 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 02:01:37,285 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 21, batch 2450, loss[loss=0.2083, simple_loss=0.287, pruned_loss=0.06479, over 8454.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2881, pruned_loss=0.06323, over 1603135.29 frames. ], batch size: 27, lr: 3.62e-03, grad_scale: 16.0 +2023-02-07 02:01:54,112 INFO [scaling.py:679] (0/4) 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] (0/4) Epoch 21, batch 2500, loss[loss=0.2045, simple_loss=0.286, pruned_loss=0.06147, over 8734.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.288, pruned_loss=0.06306, over 1603074.51 frames. ], batch size: 34, lr: 3.62e-03, grad_scale: 8.0 +2023-02-07 02:02:22,147 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) Epoch 21, batch 2550, loss[loss=0.1831, simple_loss=0.2756, pruned_loss=0.04531, over 8195.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2892, pruned_loss=0.06346, over 1611533.05 frames. ], batch size: 23, lr: 3.62e-03, grad_scale: 8.0 +2023-02-07 02:02:54,990 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9393, 2.0198, 1.7445, 2.5432, 1.0942, 1.5543, 1.8170, 1.9990], + device='cuda:0'), covar=tensor([0.0778, 0.0898, 0.0959, 0.0480, 0.1143, 0.1383, 0.0872, 0.0841], + device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0195, 0.0244, 0.0212, 0.0204, 0.0244, 0.0250, 0.0208], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 02:03:22,648 INFO [train.py:901] (0/4) Epoch 21, batch 2600, loss[loss=0.233, simple_loss=0.3147, pruned_loss=0.07568, over 8627.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2877, pruned_loss=0.06291, over 1608063.02 frames. ], batch size: 34, lr: 3.62e-03, grad_scale: 8.0 +2023-02-07 02:03:40,891 INFO [optim.py:369] (0/4) 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 02:03:56,823 INFO [train.py:901] (0/4) Epoch 21, batch 2650, loss[loss=0.2279, simple_loss=0.3119, pruned_loss=0.07196, over 8288.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2876, pruned_loss=0.06316, over 1608309.23 frames. ], batch size: 23, lr: 3.62e-03, grad_scale: 8.0 +2023-02-07 02:04:33,147 INFO [train.py:901] (0/4) Epoch 21, batch 2700, loss[loss=0.2484, simple_loss=0.3252, pruned_loss=0.08585, over 8496.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2882, pruned_loss=0.06312, over 1609898.19 frames. ], batch size: 26, lr: 3.62e-03, grad_scale: 8.0 +2023-02-07 02:04:46,924 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) Epoch 21, batch 2750, loss[loss=0.1955, simple_loss=0.2813, pruned_loss=0.05478, over 8239.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2891, pruned_loss=0.06389, over 1609754.39 frames. ], batch size: 22, lr: 3.62e-03, grad_scale: 8.0 +2023-02-07 02:05:36,819 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=164452.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:05:42,229 INFO [train.py:901] (0/4) Epoch 21, batch 2800, loss[loss=0.193, simple_loss=0.2817, pruned_loss=0.05219, over 8032.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2898, pruned_loss=0.06356, over 1614188.54 frames. ], batch size: 22, lr: 3.62e-03, grad_scale: 8.0 +2023-02-07 02:06:02,588 INFO [optim.py:369] (0/4) 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 02:06:18,052 INFO [train.py:901] (0/4) Epoch 21, batch 2850, loss[loss=0.1863, simple_loss=0.25, pruned_loss=0.06128, over 7425.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2896, pruned_loss=0.06344, over 1610822.69 frames. ], batch size: 17, lr: 3.62e-03, grad_scale: 8.0 +2023-02-07 02:06:23,426 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=164517.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:06:51,356 INFO [train.py:901] (0/4) Epoch 21, batch 2900, loss[loss=0.1758, simple_loss=0.2562, pruned_loss=0.04768, over 7928.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2905, pruned_loss=0.0642, over 1607795.15 frames. ], batch size: 20, lr: 3.62e-03, grad_scale: 8.0 +2023-02-07 02:06:56,992 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=164567.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:07:09,749 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-07 02:07:11,683 INFO [optim.py:369] (0/4) 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 02:07:11,856 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164586.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:07:18,890 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164596.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:07:28,722 INFO [train.py:901] (0/4) Epoch 21, batch 2950, loss[loss=0.189, simple_loss=0.2722, pruned_loss=0.05288, over 8616.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2886, pruned_loss=0.06339, over 1606133.75 frames. ], batch size: 31, lr: 3.62e-03, grad_scale: 8.0 +2023-02-07 02:07:44,449 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=164632.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:08:02,298 INFO [train.py:901] (0/4) Epoch 21, batch 3000, loss[loss=0.1852, simple_loss=0.2627, pruned_loss=0.05386, over 7556.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2884, pruned_loss=0.06346, over 1606088.97 frames. ], batch size: 18, lr: 3.62e-03, grad_scale: 8.0 +2023-02-07 02:08:02,299 INFO [train.py:926] (0/4) Computing validation loss +2023-02-07 02:08:15,068 INFO [train.py:935] (0/4) Epoch 21, validation: loss=0.1742, simple_loss=0.2744, pruned_loss=0.03706, over 944034.00 frames. +2023-02-07 02:08:15,069 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-07 02:08:26,759 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164676.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:08:33,565 INFO [optim.py:369] (0/4) 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 02:08:38,604 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.34 vs. limit=5.0 +2023-02-07 02:08:49,850 INFO [train.py:901] (0/4) Epoch 21, batch 3050, loss[loss=0.1737, simple_loss=0.2476, pruned_loss=0.0499, over 7201.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2882, pruned_loss=0.06319, over 1611581.48 frames. ], batch size: 16, lr: 3.62e-03, grad_scale: 8.0 +2023-02-07 02:08:59,333 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=164722.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:09:25,494 INFO [train.py:901] (0/4) Epoch 21, batch 3100, loss[loss=0.2035, simple_loss=0.2943, pruned_loss=0.05631, over 8034.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2882, pruned_loss=0.06289, over 1613877.61 frames. ], batch size: 22, lr: 3.62e-03, grad_scale: 8.0 +2023-02-07 02:09:29,014 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164764.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:09:33,853 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7649, 1.7517, 2.3140, 1.4775, 1.3045, 2.2830, 0.3302, 1.4836], + device='cuda:0'), covar=tensor([0.1719, 0.1248, 0.0387, 0.1366, 0.2843, 0.0342, 0.2363, 0.1439], + device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0194, 0.0127, 0.0221, 0.0271, 0.0133, 0.0170, 0.0190], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 02:09:43,647 INFO [optim.py:369] (0/4) 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 02:09:59,128 INFO [train.py:901] (0/4) Epoch 21, batch 3150, loss[loss=0.224, simple_loss=0.3027, pruned_loss=0.0726, over 8492.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2877, pruned_loss=0.06255, over 1612161.33 frames. ], batch size: 26, lr: 3.62e-03, grad_scale: 8.0 +2023-02-07 02:10:08,642 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164823.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:10:19,447 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=164837.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:10:22,083 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9543, 1.6217, 3.4112, 1.4277, 2.3556, 3.7548, 3.8766, 3.2405], + device='cuda:0'), covar=tensor([0.1143, 0.1768, 0.0350, 0.2214, 0.1052, 0.0230, 0.0469, 0.0531], + device='cuda:0'), in_proj_covar=tensor([0.0290, 0.0320, 0.0288, 0.0314, 0.0304, 0.0261, 0.0410, 0.0306], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 02:10:26,875 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 21, batch 3200, loss[loss=0.2212, simple_loss=0.2981, pruned_loss=0.07222, over 8078.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2891, pruned_loss=0.0632, over 1616035.84 frames. ], batch size: 21, lr: 3.61e-03, grad_scale: 8.0 +2023-02-07 02:10:54,107 INFO [optim.py:369] (0/4) 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 02:10:55,690 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164888.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:10:56,572 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-07 02:11:05,563 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.3368, 1.6933, 4.4930, 1.8145, 3.9881, 3.7636, 4.0516, 3.9522], + device='cuda:0'), covar=tensor([0.0576, 0.4436, 0.0555, 0.4081, 0.1144, 0.0975, 0.0622, 0.0657], + device='cuda:0'), in_proj_covar=tensor([0.0627, 0.0641, 0.0685, 0.0624, 0.0705, 0.0603, 0.0602, 0.0670], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 02:11:09,479 INFO [train.py:901] (0/4) Epoch 21, batch 3250, loss[loss=0.2205, simple_loss=0.3018, pruned_loss=0.06961, over 8294.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2898, pruned_loss=0.06334, over 1618324.26 frames. ], batch size: 23, lr: 3.61e-03, grad_scale: 8.0 +2023-02-07 02:11:12,436 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164913.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:11:23,940 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=164940.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:11:44,788 INFO [train.py:901] (0/4) Epoch 21, batch 3300, loss[loss=0.1774, simple_loss=0.2728, pruned_loss=0.04098, over 8262.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2884, pruned_loss=0.06216, over 1616342.56 frames. ], batch size: 24, lr: 3.61e-03, grad_scale: 8.0 +2023-02-07 02:12:05,210 INFO [optim.py:369] (0/4) 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 02:12:12,055 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7807, 2.3443, 4.3626, 1.5932, 3.2155, 2.3107, 1.9054, 3.1426], + device='cuda:0'), covar=tensor([0.1910, 0.2800, 0.0756, 0.4596, 0.1692, 0.3219, 0.2363, 0.2188], + device='cuda:0'), in_proj_covar=tensor([0.0525, 0.0600, 0.0559, 0.0640, 0.0645, 0.0595, 0.0536, 0.0637], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 02:12:20,623 INFO [train.py:901] (0/4) Epoch 21, batch 3350, loss[loss=0.2348, simple_loss=0.3094, pruned_loss=0.08013, over 8604.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2876, pruned_loss=0.0621, over 1615526.88 frames. ], batch size: 31, lr: 3.61e-03, grad_scale: 8.0 +2023-02-07 02:12:28,059 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=165020.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:12:45,356 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165045.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:12:52,364 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165055.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:12:54,956 INFO [train.py:901] (0/4) Epoch 21, batch 3400, loss[loss=0.2265, simple_loss=0.3024, pruned_loss=0.07529, over 8514.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2878, pruned_loss=0.06267, over 1611781.76 frames. ], batch size: 26, lr: 3.61e-03, grad_scale: 8.0 +2023-02-07 02:13:15,624 INFO [optim.py:369] (0/4) 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 02:13:16,553 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6025, 1.9660, 3.0695, 1.4476, 2.2856, 2.0325, 1.7502, 2.2658], + device='cuda:0'), covar=tensor([0.1794, 0.2627, 0.0870, 0.4488, 0.1928, 0.3112, 0.2202, 0.2501], + device='cuda:0'), in_proj_covar=tensor([0.0527, 0.0600, 0.0561, 0.0642, 0.0646, 0.0596, 0.0537, 0.0638], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 02:13:20,661 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=165093.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:13:31,284 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=165108.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:13:31,888 INFO [train.py:901] (0/4) Epoch 21, batch 3450, loss[loss=0.2089, simple_loss=0.281, pruned_loss=0.06846, over 7798.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2866, pruned_loss=0.06212, over 1606496.59 frames. ], batch size: 19, lr: 3.61e-03, grad_scale: 8.0 +2023-02-07 02:13:36,359 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 +2023-02-07 02:13:38,102 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=165118.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:13:49,298 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165135.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:13:53,298 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 21, batch 3500, loss[loss=0.2203, simple_loss=0.3078, pruned_loss=0.06643, over 8488.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2886, pruned_loss=0.06315, over 1606833.26 frames. ], batch size: 49, lr: 3.61e-03, grad_scale: 8.0 +2023-02-07 02:14:10,624 WARNING [train.py:1067] (0/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-07 02:14:24,648 INFO [optim.py:369] (0/4) 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 02:14:41,283 INFO [train.py:901] (0/4) Epoch 21, batch 3550, loss[loss=0.1944, simple_loss=0.2633, pruned_loss=0.06276, over 7230.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2897, pruned_loss=0.0636, over 1609480.76 frames. ], batch size: 16, lr: 3.61e-03, grad_scale: 8.0 +2023-02-07 02:14:51,650 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165223.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:15:15,602 INFO [train.py:901] (0/4) Epoch 21, batch 3600, loss[loss=0.1854, simple_loss=0.2775, pruned_loss=0.04663, over 8340.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2899, pruned_loss=0.06352, over 1609352.44 frames. ], batch size: 26, lr: 3.61e-03, grad_scale: 8.0 +2023-02-07 02:15:34,163 INFO [optim.py:369] (0/4) 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 02:15:44,472 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=165301.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:15:51,140 INFO [train.py:901] (0/4) Epoch 21, batch 3650, loss[loss=0.1772, simple_loss=0.2599, pruned_loss=0.04726, over 7551.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.289, pruned_loss=0.06317, over 1610850.67 frames. ], batch size: 18, lr: 3.61e-03, grad_scale: 8.0 +2023-02-07 02:15:52,685 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=165311.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:16:00,720 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.5551, 2.3107, 3.2674, 2.4745, 2.8981, 2.4760, 2.2180, 1.8235], + device='cuda:0'), covar=tensor([0.4991, 0.5045, 0.1751, 0.3674, 0.2668, 0.2828, 0.1918, 0.5216], + device='cuda:0'), in_proj_covar=tensor([0.0935, 0.0970, 0.0793, 0.0932, 0.0992, 0.0884, 0.0743, 0.0820], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 02:16:03,307 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=165326.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:16:10,799 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=165336.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:16:16,767 WARNING [train.py:1067] (0/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-07 02:16:25,003 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.9250, 2.4140, 3.5778, 1.9238, 1.7326, 3.5120, 0.9060, 2.2087], + device='cuda:0'), covar=tensor([0.1344, 0.1244, 0.0253, 0.1634, 0.2819, 0.0331, 0.2251, 0.1426], + device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0195, 0.0128, 0.0221, 0.0272, 0.0134, 0.0170, 0.0190], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 02:16:25,986 INFO [train.py:901] (0/4) Epoch 21, batch 3700, loss[loss=0.2558, simple_loss=0.3338, pruned_loss=0.0889, over 8325.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2888, pruned_loss=0.06313, over 1613805.86 frames. ], batch size: 25, lr: 3.61e-03, grad_scale: 8.0 +2023-02-07 02:16:43,086 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-07 02:16:44,015 INFO [optim.py:369] (0/4) 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 02:16:45,637 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6594, 1.8216, 2.7086, 1.4648, 1.9627, 1.9487, 1.7197, 1.8502], + device='cuda:0'), covar=tensor([0.1667, 0.2291, 0.0798, 0.4189, 0.1796, 0.3027, 0.2002, 0.2240], + device='cuda:0'), in_proj_covar=tensor([0.0522, 0.0595, 0.0556, 0.0636, 0.0642, 0.0591, 0.0533, 0.0632], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 02:16:47,624 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=165391.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:16:59,579 INFO [train.py:901] (0/4) Epoch 21, batch 3750, loss[loss=0.2375, simple_loss=0.3076, pruned_loss=0.08365, over 7808.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.289, pruned_loss=0.06301, over 1616074.59 frames. ], batch size: 20, lr: 3.61e-03, grad_scale: 8.0 +2023-02-07 02:17:04,499 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=165416.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:17:07,465 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 +2023-02-07 02:17:32,209 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3555, 2.0818, 2.7224, 2.2827, 2.7075, 2.3202, 2.0708, 1.4904], + device='cuda:0'), covar=tensor([0.5149, 0.4731, 0.1885, 0.3301, 0.2296, 0.2869, 0.1839, 0.4823], + device='cuda:0'), in_proj_covar=tensor([0.0932, 0.0967, 0.0790, 0.0929, 0.0988, 0.0883, 0.0739, 0.0816], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 02:17:36,650 INFO [train.py:901] (0/4) Epoch 21, batch 3800, loss[loss=0.2397, simple_loss=0.324, pruned_loss=0.07773, over 8553.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2888, pruned_loss=0.06289, over 1617021.39 frames. ], batch size: 50, lr: 3.61e-03, grad_scale: 8.0 +2023-02-07 02:17:50,387 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=165479.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:17:54,319 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 02:18:07,126 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=165504.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:18:08,528 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8502, 1.3875, 1.6171, 1.2187, 0.9203, 1.3580, 1.6014, 1.3687], + device='cuda:0'), covar=tensor([0.0547, 0.1268, 0.1733, 0.1570, 0.0653, 0.1573, 0.0699, 0.0693], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0153, 0.0190, 0.0160, 0.0099, 0.0163, 0.0113, 0.0143], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-07 02:18:10,362 INFO [train.py:901] (0/4) Epoch 21, batch 3850, loss[loss=0.2098, simple_loss=0.282, pruned_loss=0.06873, over 5868.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2885, pruned_loss=0.06245, over 1613756.79 frames. ], batch size: 13, lr: 3.61e-03, grad_scale: 8.0 +2023-02-07 02:18:15,538 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 +2023-02-07 02:18:18,549 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-07 02:18:25,783 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.9745, 1.5705, 6.1669, 2.1019, 5.6048, 5.2131, 5.6627, 5.5891], + device='cuda:0'), covar=tensor([0.0537, 0.5019, 0.0357, 0.3966, 0.0926, 0.0822, 0.0485, 0.0482], + device='cuda:0'), in_proj_covar=tensor([0.0636, 0.0652, 0.0698, 0.0635, 0.0717, 0.0614, 0.0614, 0.0683], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 02:18:46,103 INFO [train.py:901] (0/4) Epoch 21, batch 3900, loss[loss=0.1993, simple_loss=0.2906, pruned_loss=0.05396, over 8250.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2888, pruned_loss=0.06223, over 1619656.47 frames. ], batch size: 24, lr: 3.61e-03, grad_scale: 8.0 +2023-02-07 02:18:49,786 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6501, 2.3383, 4.2763, 1.5156, 3.0104, 2.3291, 1.9044, 2.9631], + device='cuda:0'), covar=tensor([0.1922, 0.2562, 0.0770, 0.4394, 0.1716, 0.3024, 0.2233, 0.2368], + device='cuda:0'), in_proj_covar=tensor([0.0522, 0.0595, 0.0555, 0.0636, 0.0642, 0.0591, 0.0533, 0.0632], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 02:18:53,040 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=165569.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:19:05,122 INFO [optim.py:369] (0/4) 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 02:19:12,742 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.5408, 1.3021, 4.7511, 1.8016, 4.2348, 3.9864, 4.2716, 4.1919], + device='cuda:0'), covar=tensor([0.0563, 0.5008, 0.0485, 0.4100, 0.1127, 0.1007, 0.0593, 0.0640], + device='cuda:0'), in_proj_covar=tensor([0.0633, 0.0648, 0.0695, 0.0632, 0.0714, 0.0612, 0.0610, 0.0681], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 02:19:14,936 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165600.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:19:20,719 INFO [train.py:901] (0/4) Epoch 21, batch 3950, loss[loss=0.226, simple_loss=0.2966, pruned_loss=0.07771, over 7812.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2882, pruned_loss=0.06172, over 1616729.83 frames. ], batch size: 20, lr: 3.61e-03, grad_scale: 8.0 +2023-02-07 02:19:55,186 INFO [train.py:901] (0/4) Epoch 21, batch 4000, loss[loss=0.1745, simple_loss=0.2516, pruned_loss=0.04872, over 7798.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2886, pruned_loss=0.06229, over 1614858.08 frames. ], batch size: 19, lr: 3.61e-03, grad_scale: 8.0 +2023-02-07 02:20:15,749 INFO [optim.py:369] (0/4) 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 02:20:31,258 INFO [train.py:901] (0/4) Epoch 21, batch 4050, loss[loss=0.2215, simple_loss=0.2934, pruned_loss=0.07482, over 8298.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2896, pruned_loss=0.06322, over 1612532.53 frames. ], batch size: 23, lr: 3.61e-03, grad_scale: 8.0 +2023-02-07 02:20:46,224 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0393, 1.8931, 3.5019, 2.4620, 2.8988, 1.9343, 1.6390, 1.7234], + device='cuda:0'), covar=tensor([0.7170, 0.6756, 0.1979, 0.3876, 0.3388, 0.4565, 0.3027, 0.5999], + device='cuda:0'), in_proj_covar=tensor([0.0939, 0.0974, 0.0797, 0.0939, 0.0993, 0.0890, 0.0747, 0.0822], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 02:21:04,769 INFO [train.py:901] (0/4) Epoch 21, batch 4100, loss[loss=0.214, simple_loss=0.2913, pruned_loss=0.06833, over 8470.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2893, pruned_loss=0.06304, over 1612872.86 frames. ], batch size: 28, lr: 3.60e-03, grad_scale: 8.0 +2023-02-07 02:21:11,135 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=165768.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:21:11,324 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.56 vs. limit=5.0 +2023-02-07 02:21:15,323 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.4097, 1.3656, 4.5779, 1.8125, 4.1066, 3.8478, 4.1301, 4.0533], + device='cuda:0'), covar=tensor([0.0570, 0.4772, 0.0463, 0.3714, 0.1016, 0.0916, 0.0604, 0.0644], + device='cuda:0'), in_proj_covar=tensor([0.0635, 0.0650, 0.0699, 0.0633, 0.0713, 0.0612, 0.0614, 0.0682], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 02:21:24,839 INFO [optim.py:369] (0/4) 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 02:21:41,894 INFO [train.py:901] (0/4) Epoch 21, batch 4150, loss[loss=0.2204, simple_loss=0.2913, pruned_loss=0.07476, over 7642.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2895, pruned_loss=0.06335, over 1614037.48 frames. ], batch size: 19, lr: 3.60e-03, grad_scale: 8.0 +2023-02-07 02:21:50,121 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=165821.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:22:13,876 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=165856.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:22:15,095 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-07 02:22:15,762 INFO [train.py:901] (0/4) Epoch 21, batch 4200, loss[loss=0.1679, simple_loss=0.256, pruned_loss=0.03989, over 7976.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2887, pruned_loss=0.06294, over 1615106.54 frames. ], batch size: 21, lr: 3.60e-03, grad_scale: 8.0 +2023-02-07 02:22:16,171 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 +2023-02-07 02:22:30,499 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=165881.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:22:33,698 INFO [optim.py:369] (0/4) 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 02:22:37,052 WARNING [train.py:1067] (0/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-07 02:22:41,244 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7510, 1.8282, 1.6534, 2.3084, 1.0666, 1.4538, 1.6499, 1.8001], + device='cuda:0'), covar=tensor([0.0820, 0.0802, 0.1007, 0.0446, 0.1166, 0.1371, 0.0829, 0.0778], + device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0199, 0.0247, 0.0215, 0.0207, 0.0249, 0.0254, 0.0211], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 02:22:50,812 INFO [train.py:901] (0/4) Epoch 21, batch 4250, loss[loss=0.2071, simple_loss=0.2831, pruned_loss=0.06557, over 7658.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2898, pruned_loss=0.06336, over 1618828.83 frames. ], batch size: 19, lr: 3.60e-03, grad_scale: 8.0 +2023-02-07 02:22:50,961 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.6974, 1.5706, 4.8898, 1.8904, 4.3531, 4.0878, 4.4017, 4.3158], + device='cuda:0'), covar=tensor([0.0561, 0.4828, 0.0412, 0.3923, 0.0988, 0.0916, 0.0603, 0.0658], + device='cuda:0'), in_proj_covar=tensor([0.0634, 0.0646, 0.0694, 0.0632, 0.0710, 0.0609, 0.0610, 0.0678], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 02:22:53,714 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=165913.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:23:26,539 INFO [train.py:901] (0/4) Epoch 21, batch 4300, loss[loss=0.2402, simple_loss=0.3319, pruned_loss=0.0743, over 8341.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2891, pruned_loss=0.06294, over 1616636.76 frames. ], batch size: 26, lr: 3.60e-03, grad_scale: 8.0 +2023-02-07 02:23:44,465 INFO [optim.py:369] (0/4) 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 02:23:54,142 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-166000.pt +2023-02-07 02:24:01,642 INFO [train.py:901] (0/4) Epoch 21, batch 4350, loss[loss=0.2142, simple_loss=0.2984, pruned_loss=0.06498, over 7980.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.29, pruned_loss=0.06311, over 1621185.59 frames. ], batch size: 21, lr: 3.60e-03, grad_scale: 8.0 +2023-02-07 02:24:11,720 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-07 02:24:15,221 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166028.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:24:30,488 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4683, 2.6896, 3.1384, 1.7570, 3.3908, 2.1948, 1.5304, 2.2205], + device='cuda:0'), covar=tensor([0.0733, 0.0338, 0.0259, 0.0723, 0.0456, 0.0721, 0.0889, 0.0499], + device='cuda:0'), in_proj_covar=tensor([0.0453, 0.0385, 0.0340, 0.0441, 0.0373, 0.0532, 0.0390, 0.0414], + device='cuda:0'), out_proj_covar=tensor([1.2177e-04, 1.0103e-04, 8.9515e-05, 1.1651e-04, 9.8250e-05, 1.5077e-04, + 1.0534e-04, 1.0995e-04], device='cuda:0') +2023-02-07 02:24:36,823 INFO [train.py:901] (0/4) Epoch 21, batch 4400, loss[loss=0.1933, simple_loss=0.287, pruned_loss=0.04984, over 7981.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2891, pruned_loss=0.06306, over 1619202.39 frames. ], batch size: 21, lr: 3.60e-03, grad_scale: 8.0 +2023-02-07 02:24:45,730 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166072.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 02:24:54,403 WARNING [train.py:1067] (0/4) 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] (0/4) 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 02:25:08,088 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9572, 2.1666, 1.7634, 2.7477, 1.3537, 1.5726, 1.9327, 2.1005], + device='cuda:0'), covar=tensor([0.0717, 0.0807, 0.0927, 0.0317, 0.1090, 0.1304, 0.0834, 0.0697], + device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0199, 0.0246, 0.0215, 0.0208, 0.0248, 0.0255, 0.0210], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 02:25:10,664 INFO [train.py:901] (0/4) Epoch 21, batch 4450, loss[loss=0.1948, simple_loss=0.2684, pruned_loss=0.06063, over 7222.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2893, pruned_loss=0.06309, over 1619554.44 frames. ], batch size: 16, lr: 3.60e-03, grad_scale: 8.0 +2023-02-07 02:25:12,781 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=166112.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:25:15,617 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8932, 3.7634, 2.0516, 2.8820, 2.5847, 1.8523, 2.5544, 3.1161], + device='cuda:0'), covar=tensor([0.1743, 0.0404, 0.1386, 0.0793, 0.0941, 0.1787, 0.1350, 0.0983], + device='cuda:0'), in_proj_covar=tensor([0.0354, 0.0234, 0.0334, 0.0308, 0.0298, 0.0333, 0.0343, 0.0315], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 02:25:31,546 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1525, 1.6247, 1.8186, 1.5288, 1.0583, 1.6364, 1.9015, 1.7113], + device='cuda:0'), covar=tensor([0.0447, 0.1182, 0.1567, 0.1371, 0.0563, 0.1386, 0.0627, 0.0607], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0152, 0.0189, 0.0159, 0.0099, 0.0162, 0.0112, 0.0143], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-07 02:25:45,428 INFO [train.py:901] (0/4) Epoch 21, batch 4500, loss[loss=0.2075, simple_loss=0.307, pruned_loss=0.05396, over 8586.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2885, pruned_loss=0.06245, over 1617557.68 frames. ], batch size: 39, lr: 3.60e-03, grad_scale: 16.0 +2023-02-07 02:25:50,235 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-07 02:25:50,295 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=166165.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:25:50,363 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166165.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:26:05,011 INFO [optim.py:369] (0/4) 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 02:26:18,169 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3291, 2.3032, 1.7156, 2.0974, 1.9528, 1.5869, 1.8579, 1.8496], + device='cuda:0'), covar=tensor([0.1663, 0.0505, 0.1379, 0.0623, 0.0757, 0.1541, 0.1137, 0.1162], + device='cuda:0'), in_proj_covar=tensor([0.0354, 0.0234, 0.0334, 0.0309, 0.0297, 0.0333, 0.0343, 0.0316], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 02:26:20,601 INFO [train.py:901] (0/4) Epoch 21, batch 4550, loss[loss=0.1861, simple_loss=0.2782, pruned_loss=0.04701, over 8033.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2888, pruned_loss=0.06224, over 1617691.21 frames. ], batch size: 22, lr: 3.60e-03, grad_scale: 16.0 +2023-02-07 02:26:33,032 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166227.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:26:54,697 INFO [train.py:901] (0/4) Epoch 21, batch 4600, loss[loss=0.1915, simple_loss=0.2802, pruned_loss=0.05135, over 8190.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2885, pruned_loss=0.06287, over 1615370.41 frames. ], batch size: 23, lr: 3.60e-03, grad_scale: 16.0 +2023-02-07 02:27:10,556 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166280.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:27:13,352 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166284.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:27:15,245 INFO [optim.py:369] (0/4) 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 02:27:30,339 INFO [train.py:901] (0/4) Epoch 21, batch 4650, loss[loss=0.1793, simple_loss=0.257, pruned_loss=0.05082, over 7422.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2883, pruned_loss=0.06272, over 1612812.90 frames. ], batch size: 17, lr: 3.60e-03, grad_scale: 16.0 +2023-02-07 02:27:30,538 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166309.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:27:41,210 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166325.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:28:03,637 INFO [train.py:901] (0/4) Epoch 21, batch 4700, loss[loss=0.1955, simple_loss=0.2648, pruned_loss=0.06313, over 7705.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2879, pruned_loss=0.06282, over 1610266.04 frames. ], batch size: 18, lr: 3.60e-03, grad_scale: 16.0 +2023-02-07 02:28:23,893 INFO [optim.py:369] (0/4) 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 02:28:30,891 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.59 vs. limit=5.0 +2023-02-07 02:28:40,135 INFO [train.py:901] (0/4) Epoch 21, batch 4750, loss[loss=0.213, simple_loss=0.3006, pruned_loss=0.06274, over 8530.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2879, pruned_loss=0.06271, over 1610594.60 frames. ], batch size: 49, lr: 3.60e-03, grad_scale: 8.0 +2023-02-07 02:28:44,972 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=166416.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 02:28:52,937 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-07 02:28:55,087 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-07 02:29:14,176 INFO [train.py:901] (0/4) Epoch 21, batch 4800, loss[loss=0.2219, simple_loss=0.2885, pruned_loss=0.07768, over 8087.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2881, pruned_loss=0.06295, over 1606096.69 frames. ], batch size: 21, lr: 3.60e-03, grad_scale: 8.0 +2023-02-07 02:29:29,969 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0338, 2.1014, 1.8441, 2.7625, 1.2895, 1.7022, 1.9442, 2.2027], + device='cuda:0'), covar=tensor([0.0735, 0.0835, 0.0926, 0.0354, 0.1142, 0.1278, 0.0951, 0.0757], + device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0197, 0.0246, 0.0215, 0.0207, 0.0245, 0.0254, 0.0210], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 02:29:30,711 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 02:29:42,736 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166499.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:29:43,986 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-07 02:29:48,838 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 21, batch 4850, loss[loss=0.221, simple_loss=0.3008, pruned_loss=0.07056, over 8363.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2887, pruned_loss=0.06261, over 1612692.56 frames. ], batch size: 26, lr: 3.60e-03, grad_scale: 8.0 +2023-02-07 02:29:49,390 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=166509.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:30:03,254 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166527.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:30:05,912 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166531.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 02:30:09,302 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166536.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:30:17,920 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166549.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:30:20,271 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.23 vs. limit=5.0 +2023-02-07 02:30:24,635 INFO [train.py:901] (0/4) Epoch 21, batch 4900, loss[loss=0.1977, simple_loss=0.2952, pruned_loss=0.05015, over 8361.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2889, pruned_loss=0.06252, over 1616021.41 frames. ], batch size: 24, lr: 3.60e-03, grad_scale: 8.0 +2023-02-07 02:30:26,189 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166561.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:30:43,060 INFO [optim.py:369] (0/4) 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 02:30:58,608 INFO [train.py:901] (0/4) Epoch 21, batch 4950, loss[loss=0.1731, simple_loss=0.2437, pruned_loss=0.05126, over 7701.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.289, pruned_loss=0.06232, over 1619860.23 frames. ], batch size: 18, lr: 3.60e-03, grad_scale: 8.0 +2023-02-07 02:31:09,512 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166624.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:31:12,171 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0504, 1.6254, 1.3627, 1.5625, 1.2989, 1.2353, 1.3014, 1.3424], + device='cuda:0'), covar=tensor([0.1168, 0.0496, 0.1336, 0.0575, 0.0816, 0.1516, 0.0972, 0.0788], + device='cuda:0'), in_proj_covar=tensor([0.0352, 0.0233, 0.0331, 0.0307, 0.0295, 0.0331, 0.0340, 0.0313], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 02:31:34,523 INFO [train.py:901] (0/4) Epoch 21, batch 5000, loss[loss=0.2008, simple_loss=0.273, pruned_loss=0.06427, over 8137.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2877, pruned_loss=0.06214, over 1612293.44 frames. ], batch size: 22, lr: 3.59e-03, grad_scale: 8.0 +2023-02-07 02:31:40,570 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6918, 1.9545, 2.0465, 1.5655, 2.1350, 1.4633, 0.6699, 1.8921], + device='cuda:0'), covar=tensor([0.0405, 0.0268, 0.0197, 0.0352, 0.0293, 0.0599, 0.0616, 0.0207], + device='cuda:0'), in_proj_covar=tensor([0.0451, 0.0385, 0.0339, 0.0441, 0.0373, 0.0531, 0.0389, 0.0413], + device='cuda:0'), out_proj_covar=tensor([1.2119e-04, 1.0106e-04, 8.9162e-05, 1.1642e-04, 9.8378e-05, 1.5046e-04, + 1.0515e-04, 1.0956e-04], device='cuda:0') +2023-02-07 02:31:41,052 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=166669.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:31:52,856 INFO [optim.py:369] (0/4) 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 02:32:07,634 INFO [train.py:901] (0/4) Epoch 21, batch 5050, loss[loss=0.1771, simple_loss=0.2578, pruned_loss=0.04826, over 7704.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2873, pruned_loss=0.06199, over 1613544.70 frames. ], batch size: 18, lr: 3.59e-03, grad_scale: 8.0 +2023-02-07 02:32:19,810 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5624, 2.5475, 1.8525, 2.3270, 2.2310, 1.6289, 2.0033, 2.1549], + device='cuda:0'), covar=tensor([0.1609, 0.0415, 0.1251, 0.0610, 0.0723, 0.1545, 0.1114, 0.1093], + device='cuda:0'), in_proj_covar=tensor([0.0352, 0.0233, 0.0332, 0.0306, 0.0295, 0.0330, 0.0341, 0.0313], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 02:32:23,024 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-07 02:32:37,034 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.8648, 1.5480, 5.9774, 2.4250, 5.3741, 5.0610, 5.4844, 5.4007], + device='cuda:0'), covar=tensor([0.0476, 0.5143, 0.0331, 0.3515, 0.0977, 0.0786, 0.0579, 0.0478], + device='cuda:0'), in_proj_covar=tensor([0.0632, 0.0643, 0.0695, 0.0629, 0.0708, 0.0607, 0.0606, 0.0679], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 02:32:38,362 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 21, batch 5100, loss[loss=0.2127, simple_loss=0.2775, pruned_loss=0.07398, over 7525.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2883, pruned_loss=0.06252, over 1614834.63 frames. ], batch size: 18, lr: 3.59e-03, grad_scale: 8.0 +2023-02-07 02:32:57,472 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9870, 1.7264, 3.3925, 1.6110, 2.4329, 3.7588, 3.8259, 3.2696], + device='cuda:0'), covar=tensor([0.1204, 0.1691, 0.0356, 0.2057, 0.1081, 0.0220, 0.0573, 0.0516], + device='cuda:0'), in_proj_covar=tensor([0.0289, 0.0318, 0.0286, 0.0312, 0.0306, 0.0262, 0.0412, 0.0304], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 02:32:59,528 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166782.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:33:00,921 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 02:33:02,974 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 21, batch 5150, loss[loss=0.197, simple_loss=0.2848, pruned_loss=0.05459, over 8515.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.2877, pruned_loss=0.06261, over 1613683.85 frames. ], batch size: 26, lr: 3.59e-03, grad_scale: 8.0 +2023-02-07 02:33:19,977 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166812.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 02:33:40,919 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=166843.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:33:52,921 INFO [train.py:901] (0/4) Epoch 21, batch 5200, loss[loss=0.2136, simple_loss=0.2966, pruned_loss=0.06533, over 8492.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2875, pruned_loss=0.06199, over 1616713.81 frames. ], batch size: 28, lr: 3.59e-03, grad_scale: 8.0 +2023-02-07 02:34:01,086 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=166871.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:34:08,930 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 02:34:17,445 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=166893.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:34:17,849 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 +2023-02-07 02:34:22,858 WARNING [train.py:1067] (0/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-07 02:34:25,852 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166905.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:34:28,017 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-02-07 02:34:28,358 INFO [train.py:901] (0/4) Epoch 21, batch 5250, loss[loss=0.1599, simple_loss=0.2359, pruned_loss=0.04192, over 7537.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2885, pruned_loss=0.06248, over 1617710.79 frames. ], batch size: 18, lr: 3.59e-03, grad_scale: 8.0 +2023-02-07 02:34:53,835 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) Epoch 21, batch 5300, loss[loss=0.1929, simple_loss=0.272, pruned_loss=0.05688, over 7653.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2889, pruned_loss=0.06265, over 1613527.73 frames. ], batch size: 19, lr: 3.59e-03, grad_scale: 8.0 +2023-02-07 02:35:06,318 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.8336, 2.2601, 3.4956, 1.7440, 1.7042, 3.4219, 0.6914, 2.1467], + device='cuda:0'), covar=tensor([0.1982, 0.1879, 0.0369, 0.2309, 0.3176, 0.0545, 0.2551, 0.1767], + device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0197, 0.0128, 0.0224, 0.0273, 0.0137, 0.0173, 0.0193], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 02:35:21,119 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 02:35:25,178 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6752, 2.0203, 2.1759, 1.2711, 2.2324, 1.5010, 0.6423, 1.9385], + device='cuda:0'), covar=tensor([0.0612, 0.0310, 0.0251, 0.0574, 0.0359, 0.0825, 0.0837, 0.0292], + device='cuda:0'), in_proj_covar=tensor([0.0449, 0.0386, 0.0338, 0.0442, 0.0374, 0.0530, 0.0388, 0.0413], + device='cuda:0'), out_proj_covar=tensor([1.2077e-04, 1.0129e-04, 8.8875e-05, 1.1674e-04, 9.8600e-05, 1.5001e-04, + 1.0490e-04, 1.0944e-04], device='cuda:0') +2023-02-07 02:35:37,481 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 21, batch 5350, loss[loss=0.2364, simple_loss=0.3104, pruned_loss=0.08121, over 8702.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2881, pruned_loss=0.06223, over 1616462.18 frames. ], batch size: 49, lr: 3.59e-03, grad_scale: 8.0 +2023-02-07 02:35:58,888 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167040.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:36:12,095 INFO [train.py:901] (0/4) Epoch 21, batch 5400, loss[loss=0.1816, simple_loss=0.2703, pruned_loss=0.04641, over 8501.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.2883, pruned_loss=0.06237, over 1615163.17 frames. ], batch size: 26, lr: 3.59e-03, grad_scale: 8.0 +2023-02-07 02:36:16,483 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167065.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:36:32,164 INFO [optim.py:369] (0/4) 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 02:36:39,937 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=167097.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:36:47,956 INFO [train.py:901] (0/4) Epoch 21, batch 5450, loss[loss=0.2099, simple_loss=0.2843, pruned_loss=0.06771, over 7548.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.288, pruned_loss=0.06208, over 1612577.94 frames. ], batch size: 18, lr: 3.59e-03, grad_scale: 8.0 +2023-02-07 02:36:55,802 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167118.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:36:58,565 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167122.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:37:01,217 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=167126.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:37:12,858 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-07 02:37:24,084 INFO [train.py:901] (0/4) Epoch 21, batch 5500, loss[loss=0.2027, simple_loss=0.2962, pruned_loss=0.05458, over 8312.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2877, pruned_loss=0.06176, over 1609104.95 frames. ], batch size: 25, lr: 3.59e-03, grad_scale: 8.0 +2023-02-07 02:37:24,243 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.6521, 1.3686, 4.8901, 1.9103, 4.3781, 4.1041, 4.4674, 4.3443], + device='cuda:0'), covar=tensor([0.0633, 0.5137, 0.0450, 0.4197, 0.1064, 0.0909, 0.0566, 0.0631], + device='cuda:0'), in_proj_covar=tensor([0.0630, 0.0642, 0.0694, 0.0629, 0.0707, 0.0605, 0.0604, 0.0676], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 02:37:36,488 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3138, 1.6659, 1.8477, 1.6136, 1.1531, 1.7308, 2.0320, 2.0870], + device='cuda:0'), covar=tensor([0.0531, 0.1210, 0.1558, 0.1362, 0.0627, 0.1379, 0.0656, 0.0527], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0152, 0.0190, 0.0159, 0.0099, 0.0163, 0.0112, 0.0143], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-07 02:37:43,653 INFO [optim.py:369] (0/4) 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 02:37:58,443 INFO [train.py:901] (0/4) Epoch 21, batch 5550, loss[loss=0.2054, simple_loss=0.2805, pruned_loss=0.06522, over 7546.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2883, pruned_loss=0.06212, over 1614409.12 frames. ], batch size: 18, lr: 3.59e-03, grad_scale: 4.0 +2023-02-07 02:38:01,310 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167212.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:38:02,551 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167214.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:38:21,110 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167239.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:38:22,439 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167241.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:38:23,121 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167242.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:38:23,321 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 +2023-02-07 02:38:34,474 INFO [train.py:901] (0/4) Epoch 21, batch 5600, loss[loss=0.2168, simple_loss=0.2933, pruned_loss=0.07013, over 8686.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2873, pruned_loss=0.06139, over 1615257.81 frames. ], batch size: 49, lr: 3.59e-03, grad_scale: 8.0 +2023-02-07 02:38:38,113 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167264.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:38:40,051 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167267.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:38:52,849 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.9412, 2.4170, 3.6131, 1.9105, 1.9348, 3.5131, 0.6772, 2.1491], + device='cuda:0'), covar=tensor([0.1369, 0.1218, 0.0255, 0.1719, 0.2500, 0.0355, 0.2288, 0.1316], + device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0195, 0.0127, 0.0222, 0.0271, 0.0135, 0.0172, 0.0191], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 02:38:54,594 INFO [optim.py:369] (0/4) 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 02:38:54,810 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167289.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:38:56,074 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=167291.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:39:08,028 INFO [train.py:901] (0/4) Epoch 21, batch 5650, loss[loss=0.1988, simple_loss=0.2919, pruned_loss=0.05287, over 8190.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2875, pruned_loss=0.06159, over 1614034.42 frames. ], batch size: 23, lr: 3.59e-03, grad_scale: 4.0 +2023-02-07 02:39:16,511 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 +2023-02-07 02:39:18,764 WARNING [train.py:1067] (0/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-07 02:39:27,013 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4663, 1.4448, 1.8115, 1.1810, 1.1359, 1.7969, 0.2381, 1.1880], + device='cuda:0'), covar=tensor([0.1638, 0.1317, 0.0439, 0.1012, 0.2720, 0.0484, 0.2177, 0.1194], + device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0193, 0.0127, 0.0220, 0.0268, 0.0134, 0.0170, 0.0190], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 02:39:44,507 INFO [train.py:901] (0/4) Epoch 21, batch 5700, loss[loss=0.2054, simple_loss=0.2811, pruned_loss=0.06488, over 7429.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.2883, pruned_loss=0.06232, over 1614006.43 frames. ], batch size: 17, lr: 3.59e-03, grad_scale: 4.0 +2023-02-07 02:40:04,669 INFO [optim.py:369] (0/4) 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 02:40:16,545 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167406.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:40:18,418 INFO [train.py:901] (0/4) Epoch 21, batch 5750, loss[loss=0.2253, simple_loss=0.3051, pruned_loss=0.07273, over 7928.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2881, pruned_loss=0.0627, over 1613755.22 frames. ], batch size: 20, lr: 3.59e-03, grad_scale: 4.0 +2023-02-07 02:40:24,268 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-07 02:40:47,692 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167450.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:40:53,628 INFO [train.py:901] (0/4) Epoch 21, batch 5800, loss[loss=0.1942, simple_loss=0.2774, pruned_loss=0.0555, over 6802.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2874, pruned_loss=0.06188, over 1611629.75 frames. ], batch size: 15, lr: 3.59e-03, grad_scale: 4.0 +2023-02-07 02:40:55,798 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=167462.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:40:59,183 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=167466.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:41:00,730 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167468.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:41:01,411 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6006, 1.9047, 2.5941, 1.5034, 1.8492, 1.9290, 1.7090, 1.8927], + device='cuda:0'), covar=tensor([0.2113, 0.2656, 0.1040, 0.4828, 0.2030, 0.3541, 0.2502, 0.2411], + device='cuda:0'), in_proj_covar=tensor([0.0520, 0.0593, 0.0552, 0.0635, 0.0637, 0.0587, 0.0528, 0.0626], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 02:41:08,617 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3053, 2.6806, 3.1974, 1.4734, 3.1701, 1.8900, 1.5957, 2.0659], + device='cuda:0'), covar=tensor([0.0825, 0.0374, 0.0282, 0.0814, 0.0630, 0.0885, 0.0873, 0.0566], + device='cuda:0'), in_proj_covar=tensor([0.0451, 0.0387, 0.0339, 0.0442, 0.0375, 0.0533, 0.0389, 0.0415], + device='cuda:0'), out_proj_covar=tensor([1.2122e-04, 1.0153e-04, 8.9186e-05, 1.1692e-04, 9.8850e-05, 1.5066e-04, + 1.0503e-04, 1.1002e-04], device='cuda:0') +2023-02-07 02:41:15,047 INFO [optim.py:369] (0/4) 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 02:41:16,947 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 +2023-02-07 02:41:18,008 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167493.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:41:20,744 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 21, batch 5850, loss[loss=0.1835, simple_loss=0.27, pruned_loss=0.04854, over 7435.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2881, pruned_loss=0.06224, over 1612620.57 frames. ], batch size: 17, lr: 3.59e-03, grad_scale: 4.0 +2023-02-07 02:41:37,844 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167522.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:42:03,764 INFO [train.py:901] (0/4) Epoch 21, batch 5900, loss[loss=0.2226, simple_loss=0.3201, pruned_loss=0.06248, over 8134.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2874, pruned_loss=0.06202, over 1611734.82 frames. ], batch size: 22, lr: 3.59e-03, grad_scale: 4.0 +2023-02-07 02:42:16,867 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167577.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:42:19,718 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167581.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:42:25,334 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4212, 2.0006, 3.8920, 1.4202, 2.8280, 2.0563, 1.5984, 2.8328], + device='cuda:0'), covar=tensor([0.2436, 0.3243, 0.0903, 0.5257, 0.2029, 0.3781, 0.2887, 0.2470], + device='cuda:0'), in_proj_covar=tensor([0.0519, 0.0592, 0.0551, 0.0635, 0.0637, 0.0587, 0.0528, 0.0626], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 02:42:25,700 INFO [optim.py:369] (0/4) 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 02:42:39,892 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2865, 1.0631, 1.3894, 1.0553, 0.7446, 1.1827, 1.2337, 1.0566], + device='cuda:0'), covar=tensor([0.0651, 0.1693, 0.2301, 0.1845, 0.0675, 0.1961, 0.0761, 0.0769], + device='cuda:0'), in_proj_covar=tensor([0.0096, 0.0151, 0.0189, 0.0158, 0.0099, 0.0162, 0.0111, 0.0142], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-07 02:42:40,390 INFO [train.py:901] (0/4) Epoch 21, batch 5950, loss[loss=0.2684, simple_loss=0.3419, pruned_loss=0.09744, over 8539.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2884, pruned_loss=0.06281, over 1615132.96 frames. ], batch size: 39, lr: 3.58e-03, grad_scale: 4.0 +2023-02-07 02:43:14,051 INFO [train.py:901] (0/4) Epoch 21, batch 6000, loss[loss=0.1807, simple_loss=0.2606, pruned_loss=0.05037, over 7539.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.288, pruned_loss=0.06247, over 1615209.60 frames. ], batch size: 18, lr: 3.58e-03, grad_scale: 8.0 +2023-02-07 02:43:14,052 INFO [train.py:926] (0/4) Computing validation loss +2023-02-07 02:43:26,400 INFO [train.py:935] (0/4) Epoch 21, validation: loss=0.174, simple_loss=0.2741, pruned_loss=0.03692, over 944034.00 frames. +2023-02-07 02:43:26,402 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-07 02:43:28,713 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167662.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:43:38,372 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.2468, 1.2377, 3.3508, 1.0675, 2.9830, 2.7822, 3.0224, 2.9788], + device='cuda:0'), covar=tensor([0.0881, 0.4471, 0.0850, 0.4566, 0.1386, 0.1223, 0.0943, 0.0938], + device='cuda:0'), in_proj_covar=tensor([0.0631, 0.0645, 0.0699, 0.0635, 0.0711, 0.0610, 0.0612, 0.0682], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 02:43:45,659 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 02:44:01,970 INFO [train.py:901] (0/4) Epoch 21, batch 6050, loss[loss=0.2038, simple_loss=0.2965, pruned_loss=0.05561, over 8627.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2867, pruned_loss=0.06163, over 1615021.03 frames. ], batch size: 31, lr: 3.58e-03, grad_scale: 8.0 +2023-02-07 02:44:22,044 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167737.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:44:38,029 INFO [train.py:901] (0/4) Epoch 21, batch 6100, loss[loss=0.1877, simple_loss=0.2835, pruned_loss=0.04598, over 8358.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2872, pruned_loss=0.06139, over 1615493.26 frames. ], batch size: 24, lr: 3.58e-03, grad_scale: 8.0 +2023-02-07 02:44:56,031 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167785.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:44:57,192 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-07 02:44:58,541 INFO [optim.py:369] (0/4) 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 02:45:02,092 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=167794.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:45:13,116 INFO [train.py:901] (0/4) Epoch 21, batch 6150, loss[loss=0.2068, simple_loss=0.2999, pruned_loss=0.05686, over 8683.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2875, pruned_loss=0.06139, over 1616750.83 frames. ], batch size: 39, lr: 3.58e-03, grad_scale: 8.0 +2023-02-07 02:45:30,430 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167833.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:45:33,094 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167837.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:45:48,361 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167858.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:45:48,814 INFO [train.py:901] (0/4) Epoch 21, batch 6200, loss[loss=0.1839, simple_loss=0.2608, pruned_loss=0.05346, over 7445.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2871, pruned_loss=0.06117, over 1615249.97 frames. ], batch size: 17, lr: 3.58e-03, grad_scale: 8.0 +2023-02-07 02:45:51,034 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167862.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:46:09,471 INFO [optim.py:369] (0/4) 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 02:46:23,402 INFO [train.py:901] (0/4) Epoch 21, batch 6250, loss[loss=0.2212, simple_loss=0.303, pruned_loss=0.06973, over 8417.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.288, pruned_loss=0.06165, over 1611233.77 frames. ], batch size: 49, lr: 3.58e-03, grad_scale: 8.0 +2023-02-07 02:46:23,624 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167909.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:46:58,909 INFO [train.py:901] (0/4) Epoch 21, batch 6300, loss[loss=0.1957, simple_loss=0.2781, pruned_loss=0.05664, over 8258.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2893, pruned_loss=0.06242, over 1614173.91 frames. ], batch size: 24, lr: 3.58e-03, grad_scale: 8.0 +2023-02-07 02:47:20,723 INFO [optim.py:369] (0/4) 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 02:47:28,255 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-168000.pt +2023-02-07 02:47:35,243 INFO [train.py:901] (0/4) Epoch 21, batch 6350, loss[loss=0.2049, simple_loss=0.292, pruned_loss=0.05891, over 8340.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2896, pruned_loss=0.06282, over 1615050.28 frames. ], batch size: 26, lr: 3.58e-03, grad_scale: 8.0 +2023-02-07 02:48:06,023 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 21, batch 6400, loss[loss=0.1845, simple_loss=0.2648, pruned_loss=0.05209, over 7707.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2891, pruned_loss=0.06239, over 1618803.11 frames. ], batch size: 18, lr: 3.58e-03, grad_scale: 8.0 +2023-02-07 02:48:16,727 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2491, 1.9786, 2.7046, 2.1816, 2.7432, 2.2363, 2.0035, 1.4098], + device='cuda:0'), covar=tensor([0.5314, 0.4892, 0.1913, 0.3788, 0.2350, 0.3079, 0.1922, 0.5305], + device='cuda:0'), in_proj_covar=tensor([0.0940, 0.0975, 0.0800, 0.0941, 0.0998, 0.0892, 0.0746, 0.0821], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 02:48:25,108 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=168081.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:48:30,482 INFO [optim.py:369] (0/4) 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 02:48:45,435 INFO [train.py:901] (0/4) Epoch 21, batch 6450, loss[loss=0.2078, simple_loss=0.2949, pruned_loss=0.06035, over 8460.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2888, pruned_loss=0.0622, over 1621864.70 frames. ], batch size: 25, lr: 3.58e-03, grad_scale: 8.0 +2023-02-07 02:48:59,381 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=168129.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:49:19,828 INFO [train.py:901] (0/4) Epoch 21, batch 6500, loss[loss=0.2073, simple_loss=0.296, pruned_loss=0.05931, over 8462.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2891, pruned_loss=0.0623, over 1622888.46 frames. ], batch size: 25, lr: 3.58e-03, grad_scale: 8.0 +2023-02-07 02:49:24,786 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 02:49:42,173 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=168190.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:49:46,107 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168196.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:49:54,741 INFO [train.py:901] (0/4) Epoch 21, batch 6550, loss[loss=0.2355, simple_loss=0.3208, pruned_loss=0.07508, over 8336.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2873, pruned_loss=0.06126, over 1620156.96 frames. ], batch size: 26, lr: 3.58e-03, grad_scale: 8.0 +2023-02-07 02:50:19,978 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168244.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:50:20,493 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-07 02:50:24,642 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=168251.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:50:29,926 INFO [train.py:901] (0/4) Epoch 21, batch 6600, loss[loss=0.2649, simple_loss=0.3311, pruned_loss=0.09934, over 7796.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2871, pruned_loss=0.06105, over 1619316.54 frames. ], batch size: 19, lr: 3.58e-03, grad_scale: 8.0 +2023-02-07 02:50:38,738 WARNING [train.py:1067] (0/4) 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] (0/4) 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 02:50:59,933 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6490, 2.1807, 4.0748, 1.5622, 3.0495, 2.2258, 1.7387, 2.8360], + device='cuda:0'), covar=tensor([0.1853, 0.2703, 0.0741, 0.4391, 0.1661, 0.3102, 0.2304, 0.2441], + device='cuda:0'), in_proj_covar=tensor([0.0523, 0.0598, 0.0555, 0.0640, 0.0642, 0.0589, 0.0531, 0.0632], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 02:51:05,096 INFO [train.py:901] (0/4) Epoch 21, batch 6650, loss[loss=0.2873, simple_loss=0.3575, pruned_loss=0.1086, over 8560.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.288, pruned_loss=0.06193, over 1619988.62 frames. ], batch size: 31, lr: 3.58e-03, grad_scale: 8.0 +2023-02-07 02:51:22,034 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.2129, 4.0807, 3.8299, 1.9249, 3.7099, 3.8145, 3.7123, 3.5555], + device='cuda:0'), covar=tensor([0.0776, 0.0632, 0.1085, 0.4739, 0.0923, 0.0806, 0.1273, 0.0822], + device='cuda:0'), in_proj_covar=tensor([0.0522, 0.0431, 0.0430, 0.0535, 0.0423, 0.0439, 0.0420, 0.0382], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 02:51:40,103 INFO [train.py:901] (0/4) Epoch 21, batch 6700, loss[loss=0.1724, simple_loss=0.249, pruned_loss=0.04784, over 7811.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2874, pruned_loss=0.06203, over 1614787.79 frames. ], batch size: 19, lr: 3.58e-03, grad_scale: 8.0 +2023-02-07 02:52:00,447 INFO [optim.py:369] (0/4) 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 02:52:05,987 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=168397.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:52:06,319 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 +2023-02-07 02:52:14,921 INFO [train.py:901] (0/4) Epoch 21, batch 6750, loss[loss=0.2219, simple_loss=0.3097, pruned_loss=0.06704, over 8433.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2881, pruned_loss=0.06282, over 1615277.13 frames. ], batch size: 29, lr: 3.58e-03, grad_scale: 8.0 +2023-02-07 02:52:45,419 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=168452.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:52:50,753 INFO [train.py:901] (0/4) Epoch 21, batch 6800, loss[loss=0.2065, simple_loss=0.2936, pruned_loss=0.0597, over 8099.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2882, pruned_loss=0.06268, over 1613950.23 frames. ], batch size: 23, lr: 3.58e-03, grad_scale: 8.0 +2023-02-07 02:52:58,530 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-07 02:53:04,360 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=168477.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:53:12,380 INFO [optim.py:369] (0/4) 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 02:53:20,381 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=168500.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:53:26,326 INFO [train.py:901] (0/4) Epoch 21, batch 6850, loss[loss=0.1657, simple_loss=0.2486, pruned_loss=0.04145, over 7693.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2871, pruned_loss=0.06201, over 1613936.01 frames. ], batch size: 18, lr: 3.58e-03, grad_scale: 8.0 +2023-02-07 02:53:28,539 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168512.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:53:37,577 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=168525.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:53:45,940 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-07 02:53:51,543 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 +2023-02-07 02:54:00,733 INFO [train.py:901] (0/4) Epoch 21, batch 6900, loss[loss=0.2093, simple_loss=0.2919, pruned_loss=0.06332, over 8102.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.2878, pruned_loss=0.06266, over 1609118.12 frames. ], batch size: 23, lr: 3.57e-03, grad_scale: 8.0 +2023-02-07 02:54:22,243 INFO [optim.py:369] (0/4) 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 02:54:26,376 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=168595.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:54:35,728 INFO [train.py:901] (0/4) Epoch 21, batch 6950, loss[loss=0.1979, simple_loss=0.2819, pruned_loss=0.05701, over 8201.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2893, pruned_loss=0.06331, over 1613150.26 frames. ], batch size: 23, lr: 3.57e-03, grad_scale: 8.0 +2023-02-07 02:54:53,388 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-07 02:54:56,312 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1816, 3.6861, 2.2877, 2.9101, 3.0255, 2.0998, 2.9265, 2.9562], + device='cuda:0'), covar=tensor([0.1496, 0.0359, 0.1245, 0.0763, 0.0662, 0.1485, 0.1005, 0.1137], + device='cuda:0'), in_proj_covar=tensor([0.0355, 0.0234, 0.0334, 0.0306, 0.0299, 0.0334, 0.0343, 0.0315], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 02:55:10,653 INFO [train.py:901] (0/4) Epoch 21, batch 7000, loss[loss=0.1684, simple_loss=0.245, pruned_loss=0.0459, over 7436.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2879, pruned_loss=0.06263, over 1609516.80 frames. ], batch size: 17, lr: 3.57e-03, grad_scale: 8.0 +2023-02-07 02:55:13,586 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9197, 2.4884, 1.9832, 2.2357, 2.2150, 1.8798, 2.1470, 2.3012], + device='cuda:0'), covar=tensor([0.1117, 0.0386, 0.0992, 0.0560, 0.0637, 0.1236, 0.0820, 0.0836], + device='cuda:0'), in_proj_covar=tensor([0.0354, 0.0233, 0.0334, 0.0305, 0.0299, 0.0334, 0.0342, 0.0315], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 02:55:17,033 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.6641, 2.3446, 3.3366, 2.6305, 3.2131, 2.5104, 2.3303, 1.9102], + device='cuda:0'), covar=tensor([0.5026, 0.4948, 0.1755, 0.3320, 0.2230, 0.2837, 0.1722, 0.5492], + device='cuda:0'), in_proj_covar=tensor([0.0940, 0.0973, 0.0803, 0.0940, 0.0997, 0.0891, 0.0745, 0.0823], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 02:55:20,991 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9422, 1.6959, 2.0656, 1.8392, 2.0081, 1.9735, 1.7809, 0.8184], + device='cuda:0'), covar=tensor([0.5642, 0.4637, 0.1873, 0.3208, 0.2256, 0.2922, 0.1830, 0.4718], + device='cuda:0'), in_proj_covar=tensor([0.0940, 0.0973, 0.0802, 0.0940, 0.0997, 0.0891, 0.0745, 0.0822], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 02:55:31,356 INFO [optim.py:369] (0/4) 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 02:55:45,696 INFO [train.py:901] (0/4) Epoch 21, batch 7050, loss[loss=0.1935, simple_loss=0.2701, pruned_loss=0.05849, over 7940.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2881, pruned_loss=0.06284, over 1607487.74 frames. ], batch size: 20, lr: 3.57e-03, grad_scale: 8.0 +2023-02-07 02:55:46,579 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168710.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:55:48,258 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.68 vs. limit=5.0 +2023-02-07 02:55:49,318 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7901, 1.7803, 2.4731, 1.4697, 1.3014, 2.3270, 0.5206, 1.4483], + device='cuda:0'), covar=tensor([0.1683, 0.1088, 0.0293, 0.1443, 0.2967, 0.0422, 0.2250, 0.1445], + device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0195, 0.0127, 0.0222, 0.0271, 0.0136, 0.0171, 0.0192], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 02:56:19,949 INFO [train.py:901] (0/4) Epoch 21, batch 7100, loss[loss=0.1888, simple_loss=0.2805, pruned_loss=0.04857, over 8239.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2884, pruned_loss=0.06268, over 1609788.23 frames. ], batch size: 24, lr: 3.57e-03, grad_scale: 8.0 +2023-02-07 02:56:26,884 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=168768.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:56:40,683 INFO [optim.py:369] (0/4) 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 02:56:43,685 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=168793.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:56:55,259 INFO [train.py:901] (0/4) Epoch 21, batch 7150, loss[loss=0.2698, simple_loss=0.3427, pruned_loss=0.09839, over 8454.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2891, pruned_loss=0.06326, over 1606218.31 frames. ], batch size: 27, lr: 3.57e-03, grad_scale: 8.0 +2023-02-07 02:57:29,818 INFO [train.py:901] (0/4) Epoch 21, batch 7200, loss[loss=0.2124, simple_loss=0.2899, pruned_loss=0.0675, over 8591.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2887, pruned_loss=0.06286, over 1609998.86 frames. ], batch size: 31, lr: 3.57e-03, grad_scale: 8.0 +2023-02-07 02:57:51,147 INFO [optim.py:369] (0/4) 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 02:57:53,932 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0691, 1.2631, 1.2342, 0.8116, 1.2608, 1.0788, 0.1348, 1.2242], + device='cuda:0'), covar=tensor([0.0382, 0.0366, 0.0335, 0.0492, 0.0377, 0.0879, 0.0765, 0.0288], + device='cuda:0'), in_proj_covar=tensor([0.0450, 0.0389, 0.0341, 0.0442, 0.0372, 0.0532, 0.0388, 0.0415], + device='cuda:0'), out_proj_covar=tensor([1.2074e-04, 1.0213e-04, 8.9754e-05, 1.1642e-04, 9.7995e-05, 1.5043e-04, + 1.0467e-04, 1.1006e-04], device='cuda:0') +2023-02-07 02:58:04,698 INFO [train.py:901] (0/4) Epoch 21, batch 7250, loss[loss=0.1913, simple_loss=0.2672, pruned_loss=0.05769, over 7704.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2891, pruned_loss=0.06349, over 1609253.02 frames. ], batch size: 18, lr: 3.57e-03, grad_scale: 8.0 +2023-02-07 02:58:40,063 INFO [train.py:901] (0/4) Epoch 21, batch 7300, loss[loss=0.1876, simple_loss=0.2735, pruned_loss=0.05081, over 8251.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2887, pruned_loss=0.06321, over 1609818.56 frames. ], batch size: 24, lr: 3.57e-03, grad_scale: 8.0 +2023-02-07 02:58:44,981 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=168985.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:58:59,434 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=168987.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 02:59:00,593 INFO [optim.py:369] (0/4) 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 02:59:02,086 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=168991.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 02:59:13,134 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 +2023-02-07 02:59:14,674 INFO [train.py:901] (0/4) Epoch 21, batch 7350, loss[loss=0.1798, simple_loss=0.2584, pruned_loss=0.05059, over 7657.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2875, pruned_loss=0.06224, over 1607527.21 frames. ], batch size: 19, lr: 3.57e-03, grad_scale: 8.0 +2023-02-07 02:59:21,355 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 +2023-02-07 02:59:35,046 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-07 02:59:48,058 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6211, 1.8758, 3.3104, 1.4278, 2.2856, 2.0781, 1.6403, 2.3841], + device='cuda:0'), covar=tensor([0.2007, 0.2841, 0.0888, 0.4742, 0.2283, 0.3250, 0.2436, 0.2588], + device='cuda:0'), in_proj_covar=tensor([0.0526, 0.0603, 0.0558, 0.0644, 0.0645, 0.0596, 0.0534, 0.0632], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 02:59:49,824 INFO [train.py:901] (0/4) Epoch 21, batch 7400, loss[loss=0.2394, simple_loss=0.3154, pruned_loss=0.0817, over 8097.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2889, pruned_loss=0.06281, over 1612002.75 frames. ], batch size: 23, lr: 3.57e-03, grad_scale: 8.0 +2023-02-07 02:59:53,393 WARNING [train.py:1067] (0/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-07 03:00:10,728 INFO [optim.py:369] (0/4) 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 03:00:19,190 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169100.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:00:25,153 INFO [train.py:901] (0/4) Epoch 21, batch 7450, loss[loss=0.1925, simple_loss=0.2858, pruned_loss=0.04958, over 8523.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2903, pruned_loss=0.06326, over 1612229.90 frames. ], batch size: 28, lr: 3.57e-03, grad_scale: 8.0 +2023-02-07 03:00:33,880 WARNING [train.py:1067] (0/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-07 03:00:34,283 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-02-07 03:00:42,569 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169134.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:01:00,744 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.5985, 2.3576, 3.2346, 2.5946, 3.1436, 2.4596, 2.3023, 1.9203], + device='cuda:0'), covar=tensor([0.4887, 0.4767, 0.1897, 0.3830, 0.2520, 0.3071, 0.1844, 0.5435], + device='cuda:0'), in_proj_covar=tensor([0.0936, 0.0969, 0.0798, 0.0937, 0.0991, 0.0885, 0.0741, 0.0817], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 03:01:01,175 INFO [train.py:901] (0/4) Epoch 21, batch 7500, loss[loss=0.2165, simple_loss=0.3037, pruned_loss=0.06463, over 8252.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2913, pruned_loss=0.0638, over 1617143.59 frames. ], batch size: 22, lr: 3.57e-03, grad_scale: 8.0 +2023-02-07 03:01:13,523 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169177.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:01:21,454 INFO [optim.py:369] (0/4) 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 03:01:35,752 INFO [train.py:901] (0/4) Epoch 21, batch 7550, loss[loss=0.2089, simple_loss=0.2775, pruned_loss=0.0702, over 7206.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2904, pruned_loss=0.06347, over 1610089.17 frames. ], batch size: 16, lr: 3.57e-03, grad_scale: 8.0 +2023-02-07 03:01:56,176 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 +2023-02-07 03:02:09,758 INFO [train.py:901] (0/4) Epoch 21, batch 7600, loss[loss=0.2054, simple_loss=0.2914, pruned_loss=0.05974, over 8101.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2912, pruned_loss=0.06395, over 1612926.76 frames. ], batch size: 23, lr: 3.57e-03, grad_scale: 8.0 +2023-02-07 03:02:18,876 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6662, 1.4912, 1.6537, 1.3434, 0.9024, 1.4706, 1.6393, 1.5309], + device='cuda:0'), covar=tensor([0.0621, 0.1236, 0.1710, 0.1487, 0.0594, 0.1558, 0.0694, 0.0644], + device='cuda:0'), in_proj_covar=tensor([0.0096, 0.0152, 0.0190, 0.0158, 0.0099, 0.0163, 0.0112, 0.0143], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-07 03:02:32,181 INFO [optim.py:369] (0/4) 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 03:02:45,878 INFO [train.py:901] (0/4) Epoch 21, batch 7650, loss[loss=0.2348, simple_loss=0.331, pruned_loss=0.06936, over 8245.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2904, pruned_loss=0.06316, over 1616722.35 frames. ], batch size: 24, lr: 3.57e-03, grad_scale: 16.0 +2023-02-07 03:03:00,457 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=169329.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:03:01,839 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 21, batch 7700, loss[loss=0.2133, simple_loss=0.2984, pruned_loss=0.06408, over 8248.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2902, pruned_loss=0.06284, over 1618577.36 frames. ], batch size: 22, lr: 3.57e-03, grad_scale: 16.0 +2023-02-07 03:03:42,196 INFO [optim.py:369] (0/4) 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 03:03:44,241 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-07 03:03:57,018 INFO [train.py:901] (0/4) Epoch 21, batch 7750, loss[loss=0.2502, simple_loss=0.3313, pruned_loss=0.08455, over 8772.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2893, pruned_loss=0.06236, over 1620596.48 frames. ], batch size: 30, lr: 3.57e-03, grad_scale: 16.0 +2023-02-07 03:04:21,904 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=169444.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:04:22,011 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169444.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:04:23,343 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169446.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 03:04:32,647 INFO [train.py:901] (0/4) Epoch 21, batch 7800, loss[loss=0.2473, simple_loss=0.3218, pruned_loss=0.08643, over 8442.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.2886, pruned_loss=0.06215, over 1616326.52 frames. ], batch size: 27, lr: 3.57e-03, grad_scale: 16.0 +2023-02-07 03:04:35,557 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5408, 1.8926, 2.9272, 1.3928, 2.1556, 1.9589, 1.6053, 2.1983], + device='cuda:0'), covar=tensor([0.1919, 0.2576, 0.0829, 0.4509, 0.1850, 0.3075, 0.2339, 0.2219], + device='cuda:0'), in_proj_covar=tensor([0.0522, 0.0599, 0.0555, 0.0637, 0.0643, 0.0591, 0.0530, 0.0630], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 03:04:45,393 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=169478.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:04:52,677 INFO [optim.py:369] (0/4) 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 03:05:06,016 INFO [train.py:901] (0/4) Epoch 21, batch 7850, loss[loss=0.2069, simple_loss=0.2874, pruned_loss=0.0632, over 8363.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.289, pruned_loss=0.06207, over 1619878.13 frames. ], batch size: 24, lr: 3.56e-03, grad_scale: 16.0 +2023-02-07 03:05:14,131 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 21, batch 7900, loss[loss=0.236, simple_loss=0.3053, pruned_loss=0.08336, over 8476.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2881, pruned_loss=0.06228, over 1614607.37 frames. ], batch size: 27, lr: 3.56e-03, grad_scale: 16.0 +2023-02-07 03:05:39,444 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169559.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:05:59,284 INFO [optim.py:369] (0/4) 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] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169593.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:06:12,856 INFO [train.py:901] (0/4) Epoch 21, batch 7950, loss[loss=0.1849, simple_loss=0.2732, pruned_loss=0.04825, over 7810.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.288, pruned_loss=0.06243, over 1610444.62 frames. ], batch size: 20, lr: 3.56e-03, grad_scale: 16.0 +2023-02-07 03:06:31,356 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169636.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:06:33,350 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169639.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:06:46,590 INFO [train.py:901] (0/4) Epoch 21, batch 8000, loss[loss=0.211, simple_loss=0.2884, pruned_loss=0.06676, over 8235.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2873, pruned_loss=0.06194, over 1612762.33 frames. ], batch size: 24, lr: 3.56e-03, grad_scale: 16.0 +2023-02-07 03:07:06,437 INFO [optim.py:369] (0/4) 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 03:07:11,984 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([0.7877, 1.1740, 0.9482, 1.1112, 0.9824, 0.8949, 0.9258, 1.0216], + device='cuda:0'), covar=tensor([0.0826, 0.0376, 0.0942, 0.0440, 0.0585, 0.1079, 0.0704, 0.0603], + device='cuda:0'), in_proj_covar=tensor([0.0355, 0.0235, 0.0333, 0.0308, 0.0300, 0.0335, 0.0343, 0.0317], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 03:07:14,030 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169700.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:07:15,395 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169702.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 03:07:19,702 INFO [train.py:901] (0/4) Epoch 21, batch 8050, loss[loss=0.1791, simple_loss=0.2471, pruned_loss=0.05556, over 7248.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2875, pruned_loss=0.06287, over 1590851.07 frames. ], batch size: 16, lr: 3.56e-03, grad_scale: 16.0 +2023-02-07 03:07:30,618 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169725.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:07:31,994 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169727.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 03:07:42,946 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/epoch-21.pt +2023-02-07 03:07:54,371 WARNING [train.py:1067] (0/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-07 03:07:58,224 INFO [train.py:901] (0/4) Epoch 22, batch 0, loss[loss=0.2043, simple_loss=0.2857, pruned_loss=0.06147, over 8741.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2857, pruned_loss=0.06147, over 8741.00 frames. ], batch size: 30, lr: 3.48e-03, grad_scale: 16.0 +2023-02-07 03:07:58,224 INFO [train.py:926] (0/4) Computing validation loss +2023-02-07 03:08:05,366 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6807, 1.8282, 1.5764, 2.1921, 1.1301, 1.5121, 1.6520, 1.7546], + device='cuda:0'), covar=tensor([0.0753, 0.0674, 0.0980, 0.0434, 0.1034, 0.1198, 0.0692, 0.0734], + device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0196, 0.0245, 0.0213, 0.0207, 0.0246, 0.0250, 0.0209], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 03:08:08,956 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2980, 1.5696, 1.6030, 1.0801, 1.6331, 1.2299, 0.3168, 1.4700], + device='cuda:0'), covar=tensor([0.0486, 0.0388, 0.0330, 0.0517, 0.0397, 0.0949, 0.0865, 0.0317], + device='cuda:0'), in_proj_covar=tensor([0.0449, 0.0387, 0.0341, 0.0438, 0.0370, 0.0529, 0.0386, 0.0416], + device='cuda:0'), out_proj_covar=tensor([1.2047e-04, 1.0158e-04, 8.9695e-05, 1.1538e-04, 9.7332e-05, 1.4925e-04, + 1.0430e-04, 1.1037e-04], device='cuda:0') +2023-02-07 03:08:09,351 INFO [train.py:935] (0/4) Epoch 22, validation: loss=0.1743, simple_loss=0.2746, pruned_loss=0.03702, over 944034.00 frames. +2023-02-07 03:08:09,352 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-07 03:08:12,908 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169747.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:08:24,241 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-07 03:08:25,065 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169765.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:08:42,194 INFO [optim.py:369] (0/4) 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 03:08:44,175 INFO [train.py:901] (0/4) Epoch 22, batch 50, loss[loss=0.2133, simple_loss=0.2922, pruned_loss=0.06725, over 8319.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.295, pruned_loss=0.06668, over 369270.41 frames. ], batch size: 25, lr: 3.48e-03, grad_scale: 16.0 +2023-02-07 03:08:54,134 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169804.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:09:01,092 WARNING [train.py:1067] (0/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-07 03:09:02,025 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169815.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:09:19,145 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169840.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:09:20,327 INFO [train.py:901] (0/4) Epoch 22, batch 100, loss[loss=0.1987, simple_loss=0.2816, pruned_loss=0.05786, over 8245.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2907, pruned_loss=0.06368, over 647993.63 frames. ], batch size: 24, lr: 3.48e-03, grad_scale: 16.0 +2023-02-07 03:09:23,126 WARNING [train.py:1067] (0/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-07 03:09:25,364 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169849.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:09:42,069 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169874.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:09:52,894 INFO [optim.py:369] (0/4) 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 03:09:55,631 INFO [train.py:901] (0/4) Epoch 22, batch 150, loss[loss=0.2222, simple_loss=0.3082, pruned_loss=0.06809, over 8343.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.289, pruned_loss=0.06271, over 859418.51 frames. ], batch size: 26, lr: 3.48e-03, grad_scale: 16.0 +2023-02-07 03:09:55,864 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169892.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:10:12,784 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169917.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:10:30,764 INFO [train.py:901] (0/4) Epoch 22, batch 200, loss[loss=0.215, simple_loss=0.2967, pruned_loss=0.06669, over 8595.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2881, pruned_loss=0.06231, over 1026301.31 frames. ], batch size: 31, lr: 3.48e-03, grad_scale: 16.0 +2023-02-07 03:10:58,685 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=169983.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:11:02,619 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 22, batch 250, loss[loss=0.2596, simple_loss=0.3319, pruned_loss=0.09363, over 8652.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2873, pruned_loss=0.06107, over 1162383.82 frames. ], batch size: 34, lr: 3.48e-03, grad_scale: 16.0 +2023-02-07 03:11:10,171 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-170000.pt +2023-02-07 03:11:17,894 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-07 03:11:26,139 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 22, batch 300, loss[loss=0.1968, simple_loss=0.2785, pruned_loss=0.0576, over 8197.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2872, pruned_loss=0.06107, over 1264285.59 frames. ], batch size: 23, lr: 3.48e-03, grad_scale: 16.0 +2023-02-07 03:11:42,593 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8024, 2.1333, 3.1929, 1.5631, 2.6643, 2.1370, 1.8762, 2.5809], + device='cuda:0'), covar=tensor([0.1791, 0.2618, 0.0927, 0.4343, 0.1708, 0.2995, 0.2225, 0.2200], + device='cuda:0'), in_proj_covar=tensor([0.0524, 0.0601, 0.0556, 0.0638, 0.0642, 0.0589, 0.0533, 0.0630], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 03:11:56,572 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170063.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:12:13,690 INFO [optim.py:369] (0/4) 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 03:12:15,163 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=170091.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:12:15,766 INFO [train.py:901] (0/4) Epoch 22, batch 350, loss[loss=0.2046, simple_loss=0.2996, pruned_loss=0.0548, over 8462.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2863, pruned_loss=0.06109, over 1336626.82 frames. ], batch size: 25, lr: 3.48e-03, grad_scale: 16.0 +2023-02-07 03:12:19,957 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170098.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:12:27,043 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=170109.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:12:34,532 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7541, 1.9325, 2.1473, 1.4770, 2.2109, 1.6713, 0.8567, 1.9335], + device='cuda:0'), covar=tensor([0.0598, 0.0364, 0.0269, 0.0539, 0.0421, 0.0752, 0.0815, 0.0304], + device='cuda:0'), in_proj_covar=tensor([0.0451, 0.0390, 0.0343, 0.0441, 0.0372, 0.0532, 0.0387, 0.0419], + device='cuda:0'), out_proj_covar=tensor([1.2114e-04, 1.0247e-04, 9.0178e-05, 1.1608e-04, 9.7872e-05, 1.5013e-04, + 1.0452e-04, 1.1113e-04], device='cuda:0') +2023-02-07 03:12:49,716 INFO [train.py:901] (0/4) Epoch 22, batch 400, loss[loss=0.2049, simple_loss=0.2991, pruned_loss=0.05532, over 8638.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2859, pruned_loss=0.0606, over 1399391.28 frames. ], batch size: 34, lr: 3.47e-03, grad_scale: 16.0 +2023-02-07 03:12:53,697 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=170148.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:12:54,639 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 +2023-02-07 03:13:22,568 INFO [optim.py:369] (0/4) 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 03:13:24,663 INFO [train.py:901] (0/4) Epoch 22, batch 450, loss[loss=0.2392, simple_loss=0.3192, pruned_loss=0.07961, over 8335.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2852, pruned_loss=0.06028, over 1446484.14 frames. ], batch size: 26, lr: 3.47e-03, grad_scale: 16.0 +2023-02-07 03:13:34,377 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170206.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:13:46,337 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170224.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:13:58,275 INFO [train.py:901] (0/4) Epoch 22, batch 500, loss[loss=0.1779, simple_loss=0.2632, pruned_loss=0.04623, over 7660.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2864, pruned_loss=0.06085, over 1487118.48 frames. ], batch size: 19, lr: 3.47e-03, grad_scale: 16.0 +2023-02-07 03:14:13,725 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 03:14:34,518 INFO [train.py:901] (0/4) Epoch 22, batch 550, loss[loss=0.2127, simple_loss=0.3024, pruned_loss=0.06147, over 8596.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2869, pruned_loss=0.06094, over 1515544.92 frames. ], batch size: 49, lr: 3.47e-03, grad_scale: 16.0 +2023-02-07 03:15:08,212 INFO [train.py:901] (0/4) Epoch 22, batch 600, loss[loss=0.1996, simple_loss=0.2742, pruned_loss=0.06249, over 7234.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2879, pruned_loss=0.06149, over 1537819.25 frames. ], batch size: 16, lr: 3.47e-03, grad_scale: 16.0 +2023-02-07 03:15:15,787 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.5273, 1.6742, 5.6518, 2.2428, 5.1243, 4.7935, 5.1999, 5.1074], + device='cuda:0'), covar=tensor([0.0393, 0.4714, 0.0340, 0.3673, 0.0863, 0.0919, 0.0459, 0.0448], + device='cuda:0'), in_proj_covar=tensor([0.0629, 0.0641, 0.0692, 0.0625, 0.0712, 0.0608, 0.0611, 0.0676], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 03:15:16,562 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=170354.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:15:27,521 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-07 03:15:34,308 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 03:15:42,757 INFO [train.py:901] (0/4) Epoch 22, batch 650, loss[loss=0.2258, simple_loss=0.3041, pruned_loss=0.07374, over 8552.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2874, pruned_loss=0.06136, over 1557185.50 frames. ], batch size: 31, lr: 3.47e-03, grad_scale: 16.0 +2023-02-07 03:15:53,430 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=170407.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:16:17,645 INFO [train.py:901] (0/4) Epoch 22, batch 700, loss[loss=0.2625, simple_loss=0.3439, pruned_loss=0.09055, over 8758.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2876, pruned_loss=0.06129, over 1569466.63 frames. ], batch size: 30, lr: 3.47e-03, grad_scale: 16.0 +2023-02-07 03:16:27,544 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 +2023-02-07 03:16:31,452 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=170462.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:16:42,960 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170479.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:16:43,710 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=170480.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:16:49,121 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) Epoch 22, batch 750, loss[loss=0.2192, simple_loss=0.3002, pruned_loss=0.06909, over 8502.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2884, pruned_loss=0.06186, over 1577476.07 frames. ], batch size: 28, lr: 3.47e-03, grad_scale: 16.0 +2023-02-07 03:17:01,714 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=170505.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:17:11,632 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=170519.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:17:13,546 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170522.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:17:14,722 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-07 03:17:23,962 WARNING [train.py:1067] (0/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-07 03:17:27,399 INFO [train.py:901] (0/4) Epoch 22, batch 800, loss[loss=0.1919, simple_loss=0.2798, pruned_loss=0.052, over 8148.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2888, pruned_loss=0.06216, over 1591653.76 frames. ], batch size: 22, lr: 3.47e-03, grad_scale: 16.0 +2023-02-07 03:17:28,967 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=170544.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:17:57,564 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170587.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:17:58,752 INFO [optim.py:369] (0/4) 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 03:18:00,801 INFO [train.py:901] (0/4) Epoch 22, batch 850, loss[loss=0.2134, simple_loss=0.3021, pruned_loss=0.06241, over 8199.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2892, pruned_loss=0.06262, over 1598574.81 frames. ], batch size: 23, lr: 3.47e-03, grad_scale: 16.0 +2023-02-07 03:18:17,662 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 +2023-02-07 03:18:36,985 INFO [train.py:901] (0/4) Epoch 22, batch 900, loss[loss=0.1979, simple_loss=0.2981, pruned_loss=0.04883, over 8508.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2891, pruned_loss=0.06179, over 1607354.53 frames. ], batch size: 26, lr: 3.47e-03, grad_scale: 16.0 +2023-02-07 03:18:57,706 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.54 vs. limit=5.0 +2023-02-07 03:19:09,372 INFO [optim.py:369] (0/4) 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 03:19:11,437 INFO [train.py:901] (0/4) Epoch 22, batch 950, loss[loss=0.2165, simple_loss=0.3074, pruned_loss=0.06284, over 8100.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2876, pruned_loss=0.06147, over 1607772.77 frames. ], batch size: 23, lr: 3.47e-03, grad_scale: 16.0 +2023-02-07 03:19:29,273 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 +2023-02-07 03:19:43,584 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 22, batch 1000, loss[loss=0.245, simple_loss=0.3236, pruned_loss=0.08322, over 7362.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.288, pruned_loss=0.06154, over 1610600.34 frames. ], batch size: 71, lr: 3.47e-03, grad_scale: 8.0 +2023-02-07 03:20:01,244 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.1809, 1.2265, 1.4797, 1.1810, 0.6949, 1.3028, 1.1667, 0.9506], + device='cuda:0'), covar=tensor([0.0624, 0.1384, 0.1777, 0.1590, 0.0637, 0.1664, 0.0760, 0.0778], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0152, 0.0190, 0.0159, 0.0099, 0.0163, 0.0112, 0.0143], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-07 03:20:12,106 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=170778.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:20:16,587 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8642, 1.5017, 3.4409, 1.5662, 2.3543, 3.7818, 3.9248, 3.1322], + device='cuda:0'), covar=tensor([0.1337, 0.1952, 0.0429, 0.2132, 0.1268, 0.0298, 0.0640, 0.0729], + device='cuda:0'), in_proj_covar=tensor([0.0295, 0.0322, 0.0285, 0.0316, 0.0309, 0.0264, 0.0418, 0.0306], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 03:20:17,095 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-07 03:20:19,772 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 22, batch 1050, loss[loss=0.2114, simple_loss=0.2795, pruned_loss=0.07162, over 7925.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2881, pruned_loss=0.06204, over 1607990.35 frames. ], batch size: 20, lr: 3.47e-03, grad_scale: 8.0 +2023-02-07 03:20:28,513 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-07 03:20:28,701 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=170803.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:20:31,481 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.02 vs. limit=5.0 +2023-02-07 03:20:41,765 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=170823.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:20:55,953 INFO [train.py:901] (0/4) Epoch 22, batch 1100, loss[loss=0.1829, simple_loss=0.2614, pruned_loss=0.0522, over 7655.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2886, pruned_loss=0.06235, over 1609673.29 frames. ], batch size: 19, lr: 3.47e-03, grad_scale: 8.0 +2023-02-07 03:21:22,157 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6421, 1.8382, 2.6397, 1.5030, 2.1686, 1.8739, 1.6849, 2.0875], + device='cuda:0'), covar=tensor([0.1472, 0.2066, 0.0671, 0.3649, 0.1416, 0.2559, 0.1908, 0.2056], + device='cuda:0'), in_proj_covar=tensor([0.0525, 0.0603, 0.0556, 0.0643, 0.0645, 0.0592, 0.0534, 0.0630], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 03:21:29,315 INFO [optim.py:369] (0/4) 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 03:21:30,679 INFO [train.py:901] (0/4) Epoch 22, batch 1150, loss[loss=0.1889, simple_loss=0.2643, pruned_loss=0.0567, over 7926.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2884, pruned_loss=0.0624, over 1606014.85 frames. ], batch size: 20, lr: 3.47e-03, grad_scale: 8.0 +2023-02-07 03:21:37,431 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-07 03:21:56,781 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=170931.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:22:01,711 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 22, batch 1200, loss[loss=0.2082, simple_loss=0.2674, pruned_loss=0.07451, over 7600.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2882, pruned_loss=0.06247, over 1603007.74 frames. ], batch size: 17, lr: 3.47e-03, grad_scale: 8.0 +2023-02-07 03:22:07,063 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 03:22:40,088 INFO [train.py:901] (0/4) Epoch 22, batch 1250, loss[loss=0.1698, simple_loss=0.251, pruned_loss=0.04434, over 7543.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2869, pruned_loss=0.06144, over 1605166.16 frames. ], batch size: 18, lr: 3.47e-03, grad_scale: 8.0 +2023-02-07 03:22:57,662 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.8233, 3.7137, 3.3970, 1.7026, 3.2764, 3.5030, 3.3865, 3.2282], + device='cuda:0'), covar=tensor([0.0951, 0.0727, 0.1189, 0.4972, 0.1025, 0.1100, 0.1409, 0.0968], + device='cuda:0'), in_proj_covar=tensor([0.0522, 0.0430, 0.0430, 0.0530, 0.0422, 0.0441, 0.0420, 0.0382], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 03:22:58,788 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 +2023-02-07 03:23:14,554 INFO [train.py:901] (0/4) Epoch 22, batch 1300, loss[loss=0.2137, simple_loss=0.3002, pruned_loss=0.06363, over 8531.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.286, pruned_loss=0.06045, over 1611376.28 frames. ], batch size: 49, lr: 3.47e-03, grad_scale: 8.0 +2023-02-07 03:23:17,483 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) Epoch 22, batch 1350, loss[loss=0.1974, simple_loss=0.2827, pruned_loss=0.05603, over 8264.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2861, pruned_loss=0.06013, over 1615094.95 frames. ], batch size: 24, lr: 3.46e-03, grad_scale: 8.0 +2023-02-07 03:24:01,671 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171110.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:24:02,358 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.7942, 1.4754, 3.9700, 1.4125, 3.4722, 3.2690, 3.5769, 3.4482], + device='cuda:0'), covar=tensor([0.0792, 0.4942, 0.0707, 0.4600, 0.1416, 0.1170, 0.0802, 0.0854], + device='cuda:0'), in_proj_covar=tensor([0.0625, 0.0637, 0.0688, 0.0620, 0.0704, 0.0604, 0.0606, 0.0672], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 03:24:23,451 INFO [train.py:901] (0/4) Epoch 22, batch 1400, loss[loss=0.1889, simple_loss=0.276, pruned_loss=0.05094, over 8477.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2866, pruned_loss=0.06054, over 1615713.66 frames. ], batch size: 29, lr: 3.46e-03, grad_scale: 8.0 +2023-02-07 03:24:23,825 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.93 vs. limit=5.0 +2023-02-07 03:24:51,098 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1759, 2.4045, 1.8897, 3.0137, 1.3789, 1.6478, 2.0699, 2.2744], + device='cuda:0'), covar=tensor([0.0657, 0.0794, 0.0891, 0.0325, 0.1120, 0.1284, 0.0850, 0.0851], + device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0197, 0.0245, 0.0215, 0.0208, 0.0248, 0.0251, 0.0212], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 03:24:55,486 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 22, batch 1450, loss[loss=0.2617, simple_loss=0.3368, pruned_loss=0.09332, over 8499.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2881, pruned_loss=0.0616, over 1612597.21 frames. ], batch size: 28, lr: 3.46e-03, grad_scale: 8.0 +2023-02-07 03:24:58,888 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171194.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:25:06,224 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-07 03:25:12,521 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171214.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:25:16,679 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=171219.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:25:32,447 INFO [train.py:901] (0/4) Epoch 22, batch 1500, loss[loss=0.2259, simple_loss=0.2979, pruned_loss=0.07689, over 7802.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2882, pruned_loss=0.06189, over 1615563.01 frames. ], batch size: 20, lr: 3.46e-03, grad_scale: 8.0 +2023-02-07 03:26:04,592 INFO [optim.py:369] (0/4) 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 03:26:04,686 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=171290.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:26:05,960 INFO [train.py:901] (0/4) Epoch 22, batch 1550, loss[loss=0.208, simple_loss=0.2898, pruned_loss=0.06307, over 8035.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2889, pruned_loss=0.06229, over 1613748.91 frames. ], batch size: 22, lr: 3.46e-03, grad_scale: 8.0 +2023-02-07 03:26:12,937 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171302.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:26:25,492 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7008, 1.8242, 1.6240, 2.2602, 0.9815, 1.4042, 1.6104, 1.8495], + device='cuda:0'), covar=tensor([0.0776, 0.0803, 0.0920, 0.0436, 0.1271, 0.1392, 0.0816, 0.0754], + device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0196, 0.0244, 0.0214, 0.0207, 0.0247, 0.0250, 0.0211], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 03:26:30,107 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=171327.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:26:37,454 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.6756, 1.4794, 4.9040, 1.7982, 4.3808, 4.1510, 4.4257, 4.3181], + device='cuda:0'), covar=tensor([0.0566, 0.4707, 0.0530, 0.3932, 0.1115, 0.0938, 0.0590, 0.0621], + device='cuda:0'), in_proj_covar=tensor([0.0629, 0.0642, 0.0693, 0.0623, 0.0708, 0.0608, 0.0611, 0.0676], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 03:26:40,692 INFO [train.py:901] (0/4) Epoch 22, batch 1600, loss[loss=0.2174, simple_loss=0.2981, pruned_loss=0.06833, over 8527.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2886, pruned_loss=0.0621, over 1613110.52 frames. ], batch size: 28, lr: 3.46e-03, grad_scale: 8.0 +2023-02-07 03:26:55,733 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171363.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:27:13,631 INFO [optim.py:369] (0/4) 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 03:27:14,996 INFO [train.py:901] (0/4) Epoch 22, batch 1650, loss[loss=0.2401, simple_loss=0.3263, pruned_loss=0.07691, over 8495.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2891, pruned_loss=0.06239, over 1615192.14 frames. ], batch size: 29, lr: 3.46e-03, grad_scale: 8.0 +2023-02-07 03:27:24,096 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171405.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:27:51,026 INFO [train.py:901] (0/4) Epoch 22, batch 1700, loss[loss=0.1678, simple_loss=0.2628, pruned_loss=0.03641, over 8242.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2882, pruned_loss=0.0618, over 1614045.42 frames. ], batch size: 22, lr: 3.46e-03, grad_scale: 8.0 +2023-02-07 03:27:59,269 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=171454.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:28:18,870 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 +2023-02-07 03:28:24,566 INFO [optim.py:369] (0/4) 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 03:28:25,936 INFO [train.py:901] (0/4) Epoch 22, batch 1750, loss[loss=0.1535, simple_loss=0.2385, pruned_loss=0.03429, over 7658.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2875, pruned_loss=0.06177, over 1614945.49 frames. ], batch size: 19, lr: 3.46e-03, grad_scale: 8.0 +2023-02-07 03:28:42,133 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171516.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:29:00,474 INFO [train.py:901] (0/4) Epoch 22, batch 1800, loss[loss=0.2009, simple_loss=0.2826, pruned_loss=0.05959, over 8651.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2867, pruned_loss=0.06158, over 1612124.50 frames. ], batch size: 34, lr: 3.46e-03, grad_scale: 8.0 +2023-02-07 03:29:11,504 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) 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] (0/4) Epoch 22, batch 1850, loss[loss=0.1638, simple_loss=0.2362, pruned_loss=0.04566, over 7425.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2867, pruned_loss=0.06172, over 1605126.42 frames. ], batch size: 17, lr: 3.46e-03, grad_scale: 8.0 +2023-02-07 03:30:10,046 INFO [train.py:901] (0/4) Epoch 22, batch 1900, loss[loss=0.1777, simple_loss=0.2648, pruned_loss=0.04528, over 8081.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2878, pruned_loss=0.06195, over 1608260.79 frames. ], batch size: 21, lr: 3.46e-03, grad_scale: 8.0 +2023-02-07 03:30:24,233 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171661.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:30:32,191 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171673.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:30:36,788 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-07 03:30:41,591 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=171686.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:30:44,065 INFO [optim.py:369] (0/4) 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 03:30:45,467 INFO [train.py:901] (0/4) Epoch 22, batch 1950, loss[loss=0.1917, simple_loss=0.2647, pruned_loss=0.05935, over 7704.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2882, pruned_loss=0.06174, over 1614133.45 frames. ], batch size: 18, lr: 3.46e-03, grad_scale: 8.0 +2023-02-07 03:30:48,015 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-07 03:30:56,295 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=171707.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:31:07,794 WARNING [train.py:1067] (0/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-07 03:31:20,037 INFO [train.py:901] (0/4) Epoch 22, batch 2000, loss[loss=0.2441, simple_loss=0.3122, pruned_loss=0.08807, over 8515.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2872, pruned_loss=0.06123, over 1617339.55 frames. ], batch size: 28, lr: 3.46e-03, grad_scale: 8.0 +2023-02-07 03:31:43,133 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0869, 1.6125, 4.3301, 1.9204, 2.3534, 4.9092, 5.0657, 4.2052], + device='cuda:0'), covar=tensor([0.1378, 0.1976, 0.0279, 0.2098, 0.1344, 0.0178, 0.0379, 0.0584], + device='cuda:0'), in_proj_covar=tensor([0.0295, 0.0323, 0.0285, 0.0318, 0.0307, 0.0264, 0.0419, 0.0305], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 03:31:53,992 INFO [optim.py:369] (0/4) 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 03:31:55,408 INFO [train.py:901] (0/4) Epoch 22, batch 2050, loss[loss=0.2926, simple_loss=0.3598, pruned_loss=0.1127, over 7322.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2873, pruned_loss=0.0612, over 1619128.13 frames. ], batch size: 72, lr: 3.46e-03, grad_scale: 8.0 +2023-02-07 03:32:17,604 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171822.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:32:19,713 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171825.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:32:31,138 INFO [train.py:901] (0/4) Epoch 22, batch 2100, loss[loss=0.1855, simple_loss=0.2608, pruned_loss=0.05509, over 7801.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2869, pruned_loss=0.06118, over 1617442.06 frames. ], batch size: 19, lr: 3.46e-03, grad_scale: 8.0 +2023-02-07 03:32:36,880 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=171860.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:32:45,790 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6619, 1.7395, 2.4917, 1.6731, 1.3114, 2.4555, 0.5821, 1.4672], + device='cuda:0'), covar=tensor([0.1955, 0.1284, 0.0342, 0.1294, 0.2857, 0.0315, 0.2322, 0.1508], + device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0198, 0.0127, 0.0223, 0.0272, 0.0137, 0.0171, 0.0194], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 03:33:05,591 INFO [optim.py:369] (0/4) 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 03:33:06,897 INFO [train.py:901] (0/4) Epoch 22, batch 2150, loss[loss=0.1927, simple_loss=0.2794, pruned_loss=0.05301, over 8517.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.288, pruned_loss=0.06268, over 1617132.30 frames. ], batch size: 49, lr: 3.46e-03, grad_scale: 8.0 +2023-02-07 03:33:33,121 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171929.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:33:42,596 INFO [train.py:901] (0/4) Epoch 22, batch 2200, loss[loss=0.2189, simple_loss=0.3015, pruned_loss=0.06812, over 7960.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2876, pruned_loss=0.06251, over 1608453.93 frames. ], batch size: 21, lr: 3.46e-03, grad_scale: 8.0 +2023-02-07 03:33:51,024 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=171954.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:33:51,678 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171955.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:34:04,375 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.53 vs. limit=2.0 +2023-02-07 03:34:05,575 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171975.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:34:15,544 INFO [optim.py:369] (0/4) 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 03:34:16,931 INFO [train.py:901] (0/4) Epoch 22, batch 2250, loss[loss=0.1585, simple_loss=0.2332, pruned_loss=0.04195, over 7704.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2863, pruned_loss=0.06196, over 1607519.27 frames. ], batch size: 18, lr: 3.46e-03, grad_scale: 8.0 +2023-02-07 03:34:17,301 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 +2023-02-07 03:34:22,551 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-172000.pt +2023-02-07 03:34:42,051 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.8046, 1.5827, 5.9933, 2.0963, 5.3289, 5.0088, 5.4722, 5.4071], + device='cuda:0'), covar=tensor([0.0563, 0.4820, 0.0333, 0.3971, 0.1135, 0.0917, 0.0645, 0.0587], + device='cuda:0'), in_proj_covar=tensor([0.0629, 0.0645, 0.0692, 0.0623, 0.0714, 0.0612, 0.0610, 0.0675], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 03:34:54,270 INFO [train.py:901] (0/4) Epoch 22, batch 2300, loss[loss=0.2005, simple_loss=0.2952, pruned_loss=0.05284, over 8245.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2864, pruned_loss=0.06173, over 1607187.82 frames. ], batch size: 24, lr: 3.46e-03, grad_scale: 8.0 +2023-02-07 03:35:20,089 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=172078.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:35:21,688 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 +2023-02-07 03:35:28,304 INFO [optim.py:369] (0/4) 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 03:35:29,626 INFO [train.py:901] (0/4) Epoch 22, batch 2350, loss[loss=0.2265, simple_loss=0.309, pruned_loss=0.07196, over 8454.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.286, pruned_loss=0.06119, over 1613360.29 frames. ], batch size: 25, lr: 3.45e-03, grad_scale: 8.0 +2023-02-07 03:35:37,297 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=172103.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:36:01,297 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=172136.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:36:05,215 INFO [train.py:901] (0/4) Epoch 22, batch 2400, loss[loss=0.2217, simple_loss=0.3013, pruned_loss=0.07109, over 8028.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2882, pruned_loss=0.06193, over 1621183.91 frames. ], batch size: 22, lr: 3.45e-03, grad_scale: 8.0 +2023-02-07 03:36:08,333 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.95 vs. limit=5.0 +2023-02-07 03:36:11,438 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9756, 1.6293, 4.2184, 1.5581, 2.6170, 4.8630, 5.1334, 3.7105], + device='cuda:0'), covar=tensor([0.1707, 0.2389, 0.0406, 0.2866, 0.1376, 0.0260, 0.0419, 0.1021], + device='cuda:0'), in_proj_covar=tensor([0.0294, 0.0324, 0.0285, 0.0317, 0.0307, 0.0264, 0.0420, 0.0306], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 03:36:14,844 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.8150, 3.7473, 3.4228, 1.7944, 3.3137, 3.4706, 3.3665, 3.2806], + device='cuda:0'), covar=tensor([0.0905, 0.0684, 0.1265, 0.4959, 0.1000, 0.1135, 0.1559, 0.0958], + device='cuda:0'), in_proj_covar=tensor([0.0529, 0.0437, 0.0435, 0.0540, 0.0430, 0.0451, 0.0429, 0.0389], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 03:36:19,756 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2883, 2.1082, 2.6837, 2.2488, 2.5988, 2.3307, 2.1263, 1.4932], + device='cuda:0'), covar=tensor([0.5729, 0.5076, 0.1994, 0.3635, 0.2641, 0.2987, 0.2043, 0.5266], + device='cuda:0'), in_proj_covar=tensor([0.0943, 0.0976, 0.0804, 0.0941, 0.0996, 0.0893, 0.0748, 0.0826], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 03:36:39,694 INFO [optim.py:369] (0/4) 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 03:36:41,128 INFO [train.py:901] (0/4) Epoch 22, batch 2450, loss[loss=0.2496, simple_loss=0.3289, pruned_loss=0.08518, over 8328.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2884, pruned_loss=0.06194, over 1619652.20 frames. ], batch size: 26, lr: 3.45e-03, grad_scale: 8.0 +2023-02-07 03:36:45,729 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.7498, 2.4642, 3.3182, 2.6607, 3.2572, 2.6583, 2.4328, 1.9107], + device='cuda:0'), covar=tensor([0.4643, 0.4923, 0.1917, 0.3514, 0.2431, 0.2917, 0.1828, 0.5334], + device='cuda:0'), in_proj_covar=tensor([0.0944, 0.0978, 0.0805, 0.0942, 0.0997, 0.0895, 0.0748, 0.0828], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 03:37:08,124 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 22, batch 2500, loss[loss=0.2535, simple_loss=0.3171, pruned_loss=0.09498, over 7152.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2888, pruned_loss=0.0617, over 1622113.41 frames. ], batch size: 71, lr: 3.45e-03, grad_scale: 8.0 +2023-02-07 03:37:26,707 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=172256.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:37:36,998 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2255, 2.5770, 2.0540, 3.6101, 1.5852, 1.8534, 2.0706, 2.5619], + device='cuda:0'), covar=tensor([0.0756, 0.0855, 0.0834, 0.0414, 0.1148, 0.1287, 0.1098, 0.0867], + device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0197, 0.0243, 0.0215, 0.0206, 0.0248, 0.0251, 0.0210], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 03:37:50,845 INFO [optim.py:369] (0/4) 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 03:37:52,246 INFO [train.py:901] (0/4) Epoch 22, batch 2550, loss[loss=0.2138, simple_loss=0.2883, pruned_loss=0.06961, over 8095.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2895, pruned_loss=0.06268, over 1617312.50 frames. ], batch size: 21, lr: 3.45e-03, grad_scale: 8.0 +2023-02-07 03:37:56,721 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=172299.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:38:25,409 INFO [train.py:901] (0/4) Epoch 22, batch 2600, loss[loss=0.2084, simple_loss=0.2874, pruned_loss=0.06468, over 8240.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2882, pruned_loss=0.06198, over 1617501.76 frames. ], batch size: 22, lr: 3.45e-03, grad_scale: 8.0 +2023-02-07 03:38:58,389 INFO [optim.py:369] (0/4) 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 03:39:00,465 INFO [train.py:901] (0/4) Epoch 22, batch 2650, loss[loss=0.2231, simple_loss=0.2981, pruned_loss=0.07403, over 8597.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2882, pruned_loss=0.06221, over 1618747.90 frames. ], batch size: 39, lr: 3.45e-03, grad_scale: 8.0 +2023-02-07 03:39:16,257 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=172414.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:39:35,339 INFO [train.py:901] (0/4) Epoch 22, batch 2700, loss[loss=0.2355, simple_loss=0.3131, pruned_loss=0.079, over 6585.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2872, pruned_loss=0.06185, over 1615692.15 frames. ], batch size: 71, lr: 3.45e-03, grad_scale: 8.0 +2023-02-07 03:40:02,834 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=172480.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:40:09,204 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 +2023-02-07 03:40:09,438 INFO [optim.py:369] (0/4) 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 03:40:10,845 INFO [train.py:901] (0/4) Epoch 22, batch 2750, loss[loss=0.1966, simple_loss=0.2681, pruned_loss=0.06259, over 7268.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2866, pruned_loss=0.06155, over 1614180.77 frames. ], batch size: 16, lr: 3.45e-03, grad_scale: 8.0 +2023-02-07 03:40:45,668 INFO [train.py:901] (0/4) Epoch 22, batch 2800, loss[loss=0.2146, simple_loss=0.2967, pruned_loss=0.06629, over 7808.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.287, pruned_loss=0.06181, over 1613680.98 frames. ], batch size: 20, lr: 3.45e-03, grad_scale: 8.0 +2023-02-07 03:41:18,227 INFO [optim.py:369] (0/4) 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 03:41:20,389 INFO [train.py:901] (0/4) Epoch 22, batch 2850, loss[loss=0.1652, simple_loss=0.2494, pruned_loss=0.04044, over 7703.00 frames. ], tot_loss[loss=0.206, simple_loss=0.288, pruned_loss=0.06204, over 1615207.94 frames. ], batch size: 18, lr: 3.45e-03, grad_scale: 8.0 +2023-02-07 03:41:23,211 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=172595.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:41:25,862 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.4463, 4.3832, 4.0073, 1.9742, 3.8427, 4.0606, 3.9567, 3.6980], + device='cuda:0'), covar=tensor([0.0665, 0.0501, 0.0939, 0.4502, 0.0906, 0.0810, 0.1239, 0.0702], + device='cuda:0'), in_proj_covar=tensor([0.0520, 0.0431, 0.0427, 0.0532, 0.0423, 0.0443, 0.0424, 0.0383], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 03:41:37,765 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=172616.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:41:44,843 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 +2023-02-07 03:41:56,011 INFO [train.py:901] (0/4) Epoch 22, batch 2900, loss[loss=0.1977, simple_loss=0.2764, pruned_loss=0.05951, over 8084.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2874, pruned_loss=0.06195, over 1613597.93 frames. ], batch size: 21, lr: 3.45e-03, grad_scale: 8.0 +2023-02-07 03:41:57,555 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=172643.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:42:15,796 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=172670.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:42:24,161 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-07 03:42:28,903 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 22, batch 2950, loss[loss=0.1787, simple_loss=0.264, pruned_loss=0.04666, over 8085.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2891, pruned_loss=0.06272, over 1615873.39 frames. ], batch size: 21, lr: 3.45e-03, grad_scale: 8.0 +2023-02-07 03:42:32,547 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=172695.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:42:56,541 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2383, 2.0070, 2.6522, 2.2493, 2.5251, 2.2634, 2.0589, 1.3787], + device='cuda:0'), covar=tensor([0.5110, 0.4327, 0.1828, 0.3246, 0.2324, 0.2852, 0.1812, 0.4847], + device='cuda:0'), in_proj_covar=tensor([0.0944, 0.0978, 0.0805, 0.0942, 0.0995, 0.0896, 0.0745, 0.0827], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 03:43:05,664 INFO [train.py:901] (0/4) Epoch 22, batch 3000, loss[loss=0.1893, simple_loss=0.2778, pruned_loss=0.05042, over 8666.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2891, pruned_loss=0.06295, over 1610513.92 frames. ], batch size: 49, lr: 3.45e-03, grad_scale: 16.0 +2023-02-07 03:43:05,665 INFO [train.py:926] (0/4) Computing validation loss +2023-02-07 03:43:17,973 INFO [train.py:935] (0/4) Epoch 22, validation: loss=0.1735, simple_loss=0.2739, pruned_loss=0.03659, over 944034.00 frames. +2023-02-07 03:43:17,975 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-07 03:43:25,621 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=172752.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:43:51,444 INFO [optim.py:369] (0/4) 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 03:43:52,758 INFO [train.py:901] (0/4) Epoch 22, batch 3050, loss[loss=0.2277, simple_loss=0.3059, pruned_loss=0.07478, over 8346.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.289, pruned_loss=0.06268, over 1615127.87 frames. ], batch size: 26, lr: 3.45e-03, grad_scale: 16.0 +2023-02-07 03:44:15,783 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1676, 2.0358, 3.4867, 2.1817, 2.7729, 3.9364, 3.9052, 3.4932], + device='cuda:0'), covar=tensor([0.1113, 0.1569, 0.0507, 0.1624, 0.1327, 0.0221, 0.0620, 0.0504], + device='cuda:0'), in_proj_covar=tensor([0.0296, 0.0325, 0.0286, 0.0318, 0.0310, 0.0266, 0.0423, 0.0309], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 03:44:26,397 INFO [train.py:901] (0/4) Epoch 22, batch 3100, loss[loss=0.2405, simple_loss=0.324, pruned_loss=0.07847, over 8456.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2896, pruned_loss=0.063, over 1616773.76 frames. ], batch size: 25, lr: 3.45e-03, grad_scale: 16.0 +2023-02-07 03:44:31,364 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1638, 2.4926, 2.8078, 1.6978, 3.1232, 1.8314, 1.5721, 2.1411], + device='cuda:0'), covar=tensor([0.0856, 0.0442, 0.0274, 0.0836, 0.0516, 0.0924, 0.0878, 0.0620], + device='cuda:0'), in_proj_covar=tensor([0.0447, 0.0386, 0.0340, 0.0440, 0.0370, 0.0528, 0.0385, 0.0412], + device='cuda:0'), 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:0') +2023-02-07 03:44:32,687 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=172851.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:44:40,606 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=172863.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:44:44,812 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1391, 2.3241, 1.9225, 2.7805, 1.6085, 1.8395, 2.1503, 2.3962], + device='cuda:0'), covar=tensor([0.0693, 0.0691, 0.0849, 0.0405, 0.0985, 0.1105, 0.0735, 0.0744], + device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0198, 0.0246, 0.0217, 0.0208, 0.0249, 0.0252, 0.0210], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 03:44:50,782 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=172876.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:44:59,805 INFO [optim.py:369] (0/4) 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 03:45:01,201 INFO [train.py:901] (0/4) Epoch 22, batch 3150, loss[loss=0.2285, simple_loss=0.3081, pruned_loss=0.0745, over 8323.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2893, pruned_loss=0.06302, over 1617130.22 frames. ], batch size: 26, lr: 3.45e-03, grad_scale: 16.0 +2023-02-07 03:45:27,821 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.31 vs. limit=5.0 +2023-02-07 03:45:29,142 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.14 vs. limit=5.0 +2023-02-07 03:45:35,472 INFO [train.py:901] (0/4) Epoch 22, batch 3200, loss[loss=0.2101, simple_loss=0.2802, pruned_loss=0.07003, over 7709.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2896, pruned_loss=0.06313, over 1617752.59 frames. ], batch size: 18, lr: 3.45e-03, grad_scale: 16.0 +2023-02-07 03:45:47,723 INFO [zipformer.py:1185] (0/4) 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] (0/4) attn_weights_entropy = tensor([2.0712, 2.3747, 1.8896, 2.8714, 1.4351, 1.6302, 2.1459, 2.2802], + device='cuda:0'), covar=tensor([0.0701, 0.0704, 0.0886, 0.0358, 0.1038, 0.1269, 0.0732, 0.0769], + device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0198, 0.0246, 0.0217, 0.0207, 0.0249, 0.0252, 0.0210], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 03:46:06,751 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=172987.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:46:07,588 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2174, 2.0288, 2.6843, 2.1821, 2.6574, 2.2891, 2.0945, 1.5337], + device='cuda:0'), covar=tensor([0.5372, 0.4800, 0.2106, 0.3975, 0.2643, 0.3056, 0.1938, 0.5365], + device='cuda:0'), in_proj_covar=tensor([0.0945, 0.0978, 0.0805, 0.0944, 0.0997, 0.0896, 0.0748, 0.0828], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 03:46:09,262 INFO [optim.py:369] (0/4) 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 03:46:10,583 INFO [train.py:901] (0/4) Epoch 22, batch 3250, loss[loss=0.1756, simple_loss=0.2569, pruned_loss=0.04711, over 7222.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2889, pruned_loss=0.06296, over 1617243.13 frames. ], batch size: 16, lr: 3.45e-03, grad_scale: 16.0 +2023-02-07 03:46:19,603 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7204, 1.9781, 2.0343, 1.4222, 2.1584, 1.6301, 0.6884, 1.8579], + device='cuda:0'), covar=tensor([0.0506, 0.0309, 0.0261, 0.0482, 0.0425, 0.0758, 0.0760, 0.0262], + device='cuda:0'), in_proj_covar=tensor([0.0448, 0.0388, 0.0342, 0.0443, 0.0372, 0.0531, 0.0387, 0.0414], + device='cuda:0'), 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:0') +2023-02-07 03:46:40,160 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1798, 1.4357, 1.6836, 1.3886, 0.9740, 1.4101, 1.8198, 1.5166], + device='cuda:0'), covar=tensor([0.0491, 0.1300, 0.1713, 0.1460, 0.0593, 0.1541, 0.0675, 0.0675], + device='cuda:0'), in_proj_covar=tensor([0.0096, 0.0152, 0.0189, 0.0158, 0.0099, 0.0163, 0.0111, 0.0142], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-07 03:46:45,374 INFO [train.py:901] (0/4) Epoch 22, batch 3300, loss[loss=0.2135, simple_loss=0.2895, pruned_loss=0.06881, over 8133.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2892, pruned_loss=0.06282, over 1616036.97 frames. ], batch size: 22, lr: 3.45e-03, grad_scale: 8.0 +2023-02-07 03:46:45,484 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.2264, 4.2288, 3.8407, 1.9841, 3.8114, 3.8135, 3.8286, 3.6120], + device='cuda:0'), covar=tensor([0.0810, 0.0571, 0.1051, 0.4686, 0.1008, 0.0971, 0.1290, 0.0894], + device='cuda:0'), in_proj_covar=tensor([0.0524, 0.0434, 0.0431, 0.0538, 0.0426, 0.0446, 0.0426, 0.0385], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 03:47:07,548 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173075.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:47:17,924 INFO [optim.py:369] (0/4) 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 03:47:18,591 INFO [train.py:901] (0/4) Epoch 22, batch 3350, loss[loss=0.2074, simple_loss=0.2889, pruned_loss=0.06301, over 8617.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2893, pruned_loss=0.06267, over 1617216.14 frames. ], batch size: 31, lr: 3.44e-03, grad_scale: 8.0 +2023-02-07 03:47:22,020 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=173096.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:47:23,439 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.4198, 4.4020, 3.9827, 2.0778, 3.8567, 4.0078, 3.9788, 3.8247], + device='cuda:0'), covar=tensor([0.0837, 0.0572, 0.1050, 0.4760, 0.1028, 0.1052, 0.1307, 0.0870], + device='cuda:0'), in_proj_covar=tensor([0.0524, 0.0434, 0.0432, 0.0536, 0.0426, 0.0446, 0.0425, 0.0385], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 03:47:26,079 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173102.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:47:43,693 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.7843, 1.5027, 5.9591, 2.2066, 5.3509, 5.0302, 5.4545, 5.3386], + device='cuda:0'), covar=tensor([0.0490, 0.4913, 0.0322, 0.3797, 0.1021, 0.0849, 0.0521, 0.0508], + device='cuda:0'), in_proj_covar=tensor([0.0625, 0.0637, 0.0686, 0.0619, 0.0701, 0.0606, 0.0604, 0.0669], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 03:47:54,983 INFO [train.py:901] (0/4) Epoch 22, batch 3400, loss[loss=0.2413, simple_loss=0.3204, pruned_loss=0.08112, over 8183.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2891, pruned_loss=0.06203, over 1622453.12 frames. ], batch size: 23, lr: 3.44e-03, grad_scale: 8.0 +2023-02-07 03:48:12,773 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173168.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:48:21,518 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.2264, 1.4521, 4.3451, 1.5849, 3.9037, 3.5918, 3.9756, 3.8362], + device='cuda:0'), covar=tensor([0.0513, 0.4107, 0.0523, 0.3688, 0.0955, 0.0931, 0.0527, 0.0576], + device='cuda:0'), in_proj_covar=tensor([0.0625, 0.0636, 0.0685, 0.0618, 0.0700, 0.0604, 0.0603, 0.0669], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 03:48:28,293 INFO [optim.py:369] (0/4) 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 03:48:28,960 INFO [train.py:901] (0/4) Epoch 22, batch 3450, loss[loss=0.2546, simple_loss=0.3374, pruned_loss=0.08588, over 8365.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.289, pruned_loss=0.06231, over 1620332.19 frames. ], batch size: 24, lr: 3.44e-03, grad_scale: 8.0 +2023-02-07 03:48:39,544 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=173207.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:48:42,427 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 22, batch 3500, loss[loss=0.2012, simple_loss=0.2919, pruned_loss=0.05528, over 8331.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2882, pruned_loss=0.06162, over 1620505.42 frames. ], batch size: 26, lr: 3.44e-03, grad_scale: 8.0 +2023-02-07 03:49:24,793 WARNING [train.py:1067] (0/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-07 03:49:38,937 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 22, batch 3550, loss[loss=0.2087, simple_loss=0.2958, pruned_loss=0.06082, over 8108.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.29, pruned_loss=0.06284, over 1620050.61 frames. ], batch size: 23, lr: 3.44e-03, grad_scale: 8.0 +2023-02-07 03:49:50,455 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.9479, 6.0345, 5.2924, 2.6613, 5.3150, 5.6750, 5.6861, 5.5546], + device='cuda:0'), covar=tensor([0.0468, 0.0318, 0.0817, 0.4133, 0.0661, 0.0715, 0.0978, 0.0548], + device='cuda:0'), in_proj_covar=tensor([0.0523, 0.0434, 0.0432, 0.0536, 0.0427, 0.0447, 0.0426, 0.0386], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 03:50:00,021 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173327.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:50:06,880 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=173331.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:50:10,306 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8701, 1.4189, 1.6741, 1.3099, 0.9735, 1.3823, 1.6644, 1.3408], + device='cuda:0'), covar=tensor([0.0568, 0.1299, 0.1667, 0.1498, 0.0616, 0.1538, 0.0703, 0.0712], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0152, 0.0189, 0.0158, 0.0099, 0.0163, 0.0111, 0.0143], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-07 03:50:14,772 INFO [train.py:901] (0/4) Epoch 22, batch 3600, loss[loss=0.1989, simple_loss=0.2844, pruned_loss=0.05674, over 8442.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2893, pruned_loss=0.0628, over 1618337.61 frames. ], batch size: 29, lr: 3.44e-03, grad_scale: 8.0 +2023-02-07 03:50:24,998 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=173356.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:50:26,357 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=173358.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:50:43,487 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=173383.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:50:48,589 INFO [optim.py:369] (0/4) 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 03:50:49,310 INFO [train.py:901] (0/4) Epoch 22, batch 3650, loss[loss=0.2009, simple_loss=0.2876, pruned_loss=0.05715, over 8246.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2898, pruned_loss=0.06279, over 1618345.81 frames. ], batch size: 22, lr: 3.44e-03, grad_scale: 8.0 +2023-02-07 03:51:23,353 INFO [train.py:901] (0/4) Epoch 22, batch 3700, loss[loss=0.1753, simple_loss=0.2622, pruned_loss=0.04418, over 8135.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2881, pruned_loss=0.06214, over 1617115.58 frames. ], batch size: 22, lr: 3.44e-03, grad_scale: 8.0 +2023-02-07 03:51:24,746 WARNING [train.py:1067] (0/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-07 03:51:42,309 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=173467.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:51:57,921 INFO [optim.py:369] (0/4) 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 03:51:58,521 INFO [train.py:901] (0/4) Epoch 22, batch 3750, loss[loss=0.2062, simple_loss=0.2877, pruned_loss=0.06231, over 8576.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2874, pruned_loss=0.06184, over 1616987.06 frames. ], batch size: 31, lr: 3.44e-03, grad_scale: 8.0 +2023-02-07 03:51:58,725 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=173492.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:52:11,078 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173509.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 03:52:12,762 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=173512.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:52:32,722 INFO [train.py:901] (0/4) Epoch 22, batch 3800, loss[loss=0.2009, simple_loss=0.277, pruned_loss=0.06239, over 7977.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2871, pruned_loss=0.0618, over 1617226.19 frames. ], batch size: 21, lr: 3.44e-03, grad_scale: 8.0 +2023-02-07 03:52:51,257 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3874, 1.6887, 1.6744, 1.1532, 1.7535, 1.3845, 0.2368, 1.5828], + device='cuda:0'), covar=tensor([0.0432, 0.0329, 0.0281, 0.0499, 0.0372, 0.0839, 0.0849, 0.0271], + device='cuda:0'), in_proj_covar=tensor([0.0452, 0.0390, 0.0344, 0.0445, 0.0376, 0.0534, 0.0390, 0.0418], + device='cuda:0'), out_proj_covar=tensor([1.2101e-04, 1.0218e-04, 9.0511e-05, 1.1720e-04, 9.8907e-05, 1.5068e-04, + 1.0519e-04, 1.1059e-04], device='cuda:0') +2023-02-07 03:52:58,729 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=173578.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:53:07,967 INFO [optim.py:369] (0/4) 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 03:53:08,687 INFO [train.py:901] (0/4) Epoch 22, batch 3850, loss[loss=0.2153, simple_loss=0.3011, pruned_loss=0.06475, over 8561.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2871, pruned_loss=0.06158, over 1618496.86 frames. ], batch size: 31, lr: 3.44e-03, grad_scale: 8.0 +2023-02-07 03:53:16,502 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=173603.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:53:30,773 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-07 03:53:33,545 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173627.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:53:36,867 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173632.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:53:43,561 INFO [train.py:901] (0/4) Epoch 22, batch 3900, loss[loss=0.182, simple_loss=0.2801, pruned_loss=0.04196, over 8731.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2871, pruned_loss=0.06119, over 1618032.54 frames. ], batch size: 30, lr: 3.44e-03, grad_scale: 8.0 +2023-02-07 03:54:02,781 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4932, 1.4553, 1.8299, 1.2794, 1.1658, 1.8073, 0.2329, 1.1718], + device='cuda:0'), covar=tensor([0.1553, 0.1346, 0.0405, 0.0985, 0.2910, 0.0439, 0.2179, 0.1309], + device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0198, 0.0128, 0.0221, 0.0269, 0.0136, 0.0171, 0.0193], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 03:54:03,323 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 03:54:17,888 INFO [train.py:901] (0/4) Epoch 22, batch 3950, loss[loss=0.2025, simple_loss=0.2813, pruned_loss=0.0618, over 7445.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2862, pruned_loss=0.06087, over 1614270.29 frames. ], batch size: 17, lr: 3.44e-03, grad_scale: 8.0 +2023-02-07 03:54:52,805 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5167, 1.4924, 1.8296, 1.2464, 1.2027, 1.8079, 0.2673, 1.1119], + device='cuda:0'), covar=tensor([0.1578, 0.1243, 0.0380, 0.0962, 0.2543, 0.0417, 0.1944, 0.1310], + device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0197, 0.0127, 0.0220, 0.0267, 0.0136, 0.0170, 0.0193], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 03:54:53,285 INFO [train.py:901] (0/4) Epoch 22, batch 4000, loss[loss=0.2223, simple_loss=0.3026, pruned_loss=0.07106, over 8480.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2859, pruned_loss=0.06082, over 1614187.52 frames. ], batch size: 29, lr: 3.44e-03, grad_scale: 8.0 +2023-02-07 03:55:09,380 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0192, 2.2776, 1.8395, 2.7664, 1.2945, 1.5895, 1.9988, 2.2084], + device='cuda:0'), covar=tensor([0.0707, 0.0697, 0.0923, 0.0362, 0.1134, 0.1335, 0.0836, 0.0754], + device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0199, 0.0246, 0.0216, 0.0208, 0.0248, 0.0252, 0.0210], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 03:55:23,153 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173786.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:55:24,452 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0504, 1.2302, 1.1206, 1.9298, 0.8475, 1.0849, 1.5534, 1.4186], + device='cuda:0'), covar=tensor([0.1556, 0.1090, 0.1923, 0.0551, 0.1195, 0.1797, 0.0688, 0.0778], + device='cuda:0'), in_proj_covar=tensor([0.0235, 0.0200, 0.0248, 0.0217, 0.0209, 0.0250, 0.0254, 0.0211], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 03:55:26,125 INFO [optim.py:369] (0/4) 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 03:55:26,787 INFO [train.py:901] (0/4) Epoch 22, batch 4050, loss[loss=0.2299, simple_loss=0.3062, pruned_loss=0.07676, over 8461.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2874, pruned_loss=0.06151, over 1614757.30 frames. ], batch size: 25, lr: 3.44e-03, grad_scale: 8.0 +2023-02-07 03:56:02,760 INFO [train.py:901] (0/4) Epoch 22, batch 4100, loss[loss=0.2241, simple_loss=0.315, pruned_loss=0.06661, over 8480.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2889, pruned_loss=0.06239, over 1619411.89 frames. ], batch size: 29, lr: 3.44e-03, grad_scale: 8.0 +2023-02-07 03:56:10,365 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=173853.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 03:56:31,257 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=173883.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:56:36,360 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 22, batch 4150, loss[loss=0.239, simple_loss=0.3172, pruned_loss=0.08037, over 8506.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2881, pruned_loss=0.06195, over 1617925.70 frames. ], batch size: 26, lr: 3.44e-03, grad_scale: 8.0 +2023-02-07 03:56:47,600 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=173908.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:57:02,474 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-02-07 03:57:12,184 INFO [train.py:901] (0/4) Epoch 22, batch 4200, loss[loss=0.1512, simple_loss=0.2236, pruned_loss=0.03944, over 7423.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2873, pruned_loss=0.06152, over 1614328.00 frames. ], batch size: 17, lr: 3.44e-03, grad_scale: 8.0 +2023-02-07 03:57:28,909 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-07 03:57:29,741 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173968.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 03:57:35,178 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=173976.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:57:46,762 INFO [optim.py:369] (0/4) 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 03:57:46,787 INFO [train.py:901] (0/4) Epoch 22, batch 4250, loss[loss=0.2217, simple_loss=0.2808, pruned_loss=0.08127, over 7260.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2882, pruned_loss=0.06183, over 1617034.10 frames. ], batch size: 16, lr: 3.44e-03, grad_scale: 4.0 +2023-02-07 03:57:52,302 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-174000.pt +2023-02-07 03:57:55,806 WARNING [train.py:1067] (0/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-07 03:58:04,794 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.1408, 4.0962, 3.7886, 2.0569, 3.6617, 3.7440, 3.7883, 3.5746], + device='cuda:0'), covar=tensor([0.0774, 0.0597, 0.1011, 0.4517, 0.0847, 0.0997, 0.1225, 0.0853], + device='cuda:0'), in_proj_covar=tensor([0.0527, 0.0436, 0.0433, 0.0538, 0.0425, 0.0448, 0.0428, 0.0388], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 03:58:15,725 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174033.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:58:17,067 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 22, batch 4300, loss[loss=0.1959, simple_loss=0.2891, pruned_loss=0.05128, over 8330.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2883, pruned_loss=0.06169, over 1618155.27 frames. ], batch size: 26, lr: 3.44e-03, grad_scale: 4.0 +2023-02-07 03:58:21,839 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174042.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:58:25,982 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-02-07 03:58:40,389 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174067.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 03:58:56,900 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 03:58:57,400 INFO [train.py:901] (0/4) Epoch 22, batch 4350, loss[loss=0.2159, simple_loss=0.3013, pruned_loss=0.06527, over 8299.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2887, pruned_loss=0.06189, over 1619034.61 frames. ], batch size: 23, lr: 3.43e-03, grad_scale: 4.0 +2023-02-07 03:59:25,038 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-07 03:59:32,547 INFO [train.py:901] (0/4) Epoch 22, batch 4400, loss[loss=0.2151, simple_loss=0.3001, pruned_loss=0.06505, over 8466.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2902, pruned_loss=0.06254, over 1620301.58 frames. ], batch size: 25, lr: 3.43e-03, grad_scale: 8.0 +2023-02-07 03:59:36,057 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7493, 1.8556, 1.5796, 2.3291, 1.0263, 1.3948, 1.6979, 1.8009], + device='cuda:0'), covar=tensor([0.0805, 0.0745, 0.0963, 0.0406, 0.1110, 0.1363, 0.0731, 0.0796], + device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0199, 0.0246, 0.0216, 0.0208, 0.0247, 0.0250, 0.0209], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 03:59:42,802 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174157.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:00:06,461 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-07 04:00:07,758 INFO [optim.py:369] (0/4) 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 04:00:07,778 INFO [train.py:901] (0/4) Epoch 22, batch 4450, loss[loss=0.1801, simple_loss=0.2648, pruned_loss=0.04773, over 7923.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2911, pruned_loss=0.06295, over 1620422.65 frames. ], batch size: 20, lr: 3.43e-03, grad_scale: 8.0 +2023-02-07 04:00:24,477 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 +2023-02-07 04:00:26,069 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4280, 1.6244, 2.2143, 1.3203, 1.6011, 1.6934, 1.5161, 1.6222], + device='cuda:0'), covar=tensor([0.1990, 0.2674, 0.0975, 0.4753, 0.1915, 0.3478, 0.2391, 0.2090], + device='cuda:0'), in_proj_covar=tensor([0.0528, 0.0605, 0.0559, 0.0645, 0.0646, 0.0591, 0.0536, 0.0630], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 04:00:30,136 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174224.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 04:00:41,754 INFO [train.py:901] (0/4) Epoch 22, batch 4500, loss[loss=0.2372, simple_loss=0.3247, pruned_loss=0.07488, over 8635.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2908, pruned_loss=0.06332, over 1618429.90 frames. ], batch size: 31, lr: 3.43e-03, grad_scale: 8.0 +2023-02-07 04:00:46,559 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174249.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 04:00:56,984 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-07 04:01:17,042 INFO [optim.py:369] (0/4) 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 04:01:17,062 INFO [train.py:901] (0/4) Epoch 22, batch 4550, loss[loss=0.1821, simple_loss=0.2632, pruned_loss=0.05047, over 7921.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2906, pruned_loss=0.06317, over 1617487.30 frames. ], batch size: 20, lr: 3.43e-03, grad_scale: 8.0 +2023-02-07 04:01:23,301 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174301.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:01:28,566 INFO [zipformer.py:1185] (0/4) 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] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.50 vs. limit=2.0 +2023-02-07 04:01:51,063 INFO [train.py:901] (0/4) Epoch 22, batch 4600, loss[loss=0.198, simple_loss=0.2946, pruned_loss=0.05066, over 8360.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2908, pruned_loss=0.06335, over 1615441.54 frames. ], batch size: 24, lr: 3.43e-03, grad_scale: 8.0 +2023-02-07 04:01:54,749 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174347.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:02:12,049 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174372.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:02:15,437 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=174377.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:02:16,887 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=174379.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:02:25,990 INFO [optim.py:369] (0/4) 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 04:02:26,016 INFO [train.py:901] (0/4) Epoch 22, batch 4650, loss[loss=0.197, simple_loss=0.2969, pruned_loss=0.04857, over 8243.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2898, pruned_loss=0.06271, over 1618692.88 frames. ], batch size: 24, lr: 3.43e-03, grad_scale: 8.0 +2023-02-07 04:02:31,944 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 +2023-02-07 04:02:37,948 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174406.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:03:02,520 INFO [train.py:901] (0/4) Epoch 22, batch 4700, loss[loss=0.1973, simple_loss=0.2679, pruned_loss=0.06335, over 7711.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2881, pruned_loss=0.06166, over 1618320.11 frames. ], batch size: 18, lr: 3.43e-03, grad_scale: 8.0 +2023-02-07 04:03:21,309 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7518, 1.9520, 2.0965, 1.3756, 2.2165, 1.5640, 0.6917, 1.9680], + device='cuda:0'), covar=tensor([0.0557, 0.0343, 0.0267, 0.0564, 0.0359, 0.0710, 0.0849, 0.0294], + device='cuda:0'), in_proj_covar=tensor([0.0449, 0.0387, 0.0343, 0.0442, 0.0372, 0.0528, 0.0385, 0.0415], + device='cuda:0'), 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:0') +2023-02-07 04:03:37,044 INFO [optim.py:369] (0/4) 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 04:03:37,065 INFO [train.py:901] (0/4) Epoch 22, batch 4750, loss[loss=0.2085, simple_loss=0.2982, pruned_loss=0.05939, over 8254.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2887, pruned_loss=0.06201, over 1615786.50 frames. ], batch size: 24, lr: 3.43e-03, grad_scale: 8.0 +2023-02-07 04:03:37,252 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174492.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:03:38,601 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174494.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:03:43,310 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=174501.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:03:59,881 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.7219, 2.2258, 3.4116, 1.6867, 1.6840, 3.2971, 0.7426, 1.9864], + device='cuda:0'), covar=tensor([0.1548, 0.1212, 0.0241, 0.1924, 0.2634, 0.0278, 0.2172, 0.1579], + device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0198, 0.0128, 0.0220, 0.0269, 0.0136, 0.0171, 0.0194], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 04:04:04,453 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-07 04:04:06,501 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-07 04:04:12,378 INFO [train.py:901] (0/4) Epoch 22, batch 4800, loss[loss=0.2124, simple_loss=0.2988, pruned_loss=0.06299, over 8194.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2883, pruned_loss=0.06212, over 1615388.16 frames. ], batch size: 23, lr: 3.43e-03, grad_scale: 8.0 +2023-02-07 04:04:19,976 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.95 vs. limit=5.0 +2023-02-07 04:04:22,156 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.7111, 2.1572, 3.4877, 1.8474, 1.5293, 3.3928, 0.6494, 2.0184], + device='cuda:0'), covar=tensor([0.1510, 0.1385, 0.0284, 0.1918, 0.3283, 0.0315, 0.2444, 0.1639], + device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0198, 0.0129, 0.0221, 0.0270, 0.0137, 0.0172, 0.0195], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 04:04:27,173 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 +2023-02-07 04:04:46,103 INFO [optim.py:369] (0/4) 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 04:04:46,130 INFO [train.py:901] (0/4) Epoch 22, batch 4850, loss[loss=0.1992, simple_loss=0.2772, pruned_loss=0.06057, over 7538.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2896, pruned_loss=0.06304, over 1614908.45 frames. ], batch size: 18, lr: 3.43e-03, grad_scale: 8.0 +2023-02-07 04:04:55,425 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-07 04:05:02,299 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174616.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:05:20,319 INFO [train.py:901] (0/4) Epoch 22, batch 4900, loss[loss=0.1888, simple_loss=0.2761, pruned_loss=0.05081, over 8092.00 frames. ], tot_loss[loss=0.2086, simple_loss=0.2898, pruned_loss=0.06369, over 1617098.49 frames. ], batch size: 21, lr: 3.43e-03, grad_scale: 8.0 +2023-02-07 04:05:23,132 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=174645.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:05:29,307 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=174653.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:05:56,489 INFO [optim.py:369] (0/4) 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 04:05:56,510 INFO [train.py:901] (0/4) Epoch 22, batch 4950, loss[loss=0.197, simple_loss=0.2861, pruned_loss=0.054, over 8031.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2895, pruned_loss=0.06309, over 1619452.30 frames. ], batch size: 22, lr: 3.43e-03, grad_scale: 8.0 +2023-02-07 04:06:17,783 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174723.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:06:21,265 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4877, 2.4434, 1.6842, 2.2466, 1.9750, 1.3586, 1.9122, 2.0570], + device='cuda:0'), covar=tensor([0.1619, 0.0418, 0.1393, 0.0659, 0.0916, 0.1771, 0.1129, 0.1002], + device='cuda:0'), in_proj_covar=tensor([0.0350, 0.0231, 0.0331, 0.0305, 0.0297, 0.0336, 0.0339, 0.0316], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 04:06:27,422 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.6569, 2.9724, 2.6369, 4.1475, 1.8016, 2.2550, 2.7239, 3.1189], + device='cuda:0'), covar=tensor([0.0617, 0.0713, 0.0675, 0.0238, 0.1084, 0.1135, 0.0842, 0.0714], + device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0198, 0.0245, 0.0216, 0.0208, 0.0248, 0.0250, 0.0210], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 04:06:30,646 INFO [train.py:901] (0/4) Epoch 22, batch 5000, loss[loss=0.2119, simple_loss=0.2944, pruned_loss=0.06466, over 8026.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2873, pruned_loss=0.06176, over 1617330.75 frames. ], batch size: 22, lr: 3.43e-03, grad_scale: 8.0 +2023-02-07 04:06:35,000 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174748.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:06:36,235 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=174750.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:06:36,386 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174750.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:06:43,936 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174760.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:06:51,010 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174768.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:06:54,576 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174773.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:06:55,980 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174775.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:07:07,683 INFO [optim.py:369] (0/4) 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 04:07:07,703 INFO [train.py:901] (0/4) Epoch 22, batch 5050, loss[loss=0.2097, simple_loss=0.2967, pruned_loss=0.0613, over 8542.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2884, pruned_loss=0.06196, over 1626100.90 frames. ], batch size: 31, lr: 3.43e-03, grad_scale: 8.0 +2023-02-07 04:07:16,964 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0572, 1.4876, 1.7067, 1.4323, 1.0363, 1.4631, 1.9371, 1.6214], + device='cuda:0'), covar=tensor([0.0524, 0.1246, 0.1637, 0.1423, 0.0594, 0.1472, 0.0654, 0.0670], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0153, 0.0190, 0.0159, 0.0100, 0.0164, 0.0112, 0.0144], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-07 04:07:27,846 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2538, 1.9751, 2.6261, 2.1536, 2.4996, 2.2679, 2.0363, 1.4280], + device='cuda:0'), covar=tensor([0.5193, 0.4842, 0.1982, 0.3634, 0.2383, 0.2883, 0.1820, 0.5135], + device='cuda:0'), in_proj_covar=tensor([0.0935, 0.0975, 0.0800, 0.0941, 0.0989, 0.0888, 0.0745, 0.0821], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 04:07:34,469 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-07 04:07:42,672 INFO [train.py:901] (0/4) Epoch 22, batch 5100, loss[loss=0.2179, simple_loss=0.3108, pruned_loss=0.06246, over 8550.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2878, pruned_loss=0.06127, over 1626387.45 frames. ], batch size: 31, lr: 3.43e-03, grad_scale: 8.0 +2023-02-07 04:07:58,165 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174865.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:08:03,627 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174872.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:08:17,810 INFO [optim.py:369] (0/4) 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 04:08:17,830 INFO [train.py:901] (0/4) Epoch 22, batch 5150, loss[loss=0.1801, simple_loss=0.255, pruned_loss=0.05266, over 7700.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.289, pruned_loss=0.06243, over 1624507.09 frames. ], batch size: 18, lr: 3.43e-03, grad_scale: 8.0 +2023-02-07 04:08:21,121 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174897.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:08:52,061 INFO [train.py:901] (0/4) Epoch 22, batch 5200, loss[loss=0.1733, simple_loss=0.244, pruned_loss=0.05131, over 7715.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2881, pruned_loss=0.06219, over 1618252.58 frames. ], batch size: 18, lr: 3.43e-03, grad_scale: 8.0 +2023-02-07 04:08:52,890 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.8341, 1.2366, 3.9517, 1.4490, 3.4713, 3.2943, 3.6199, 3.4565], + device='cuda:0'), covar=tensor([0.0615, 0.5168, 0.0662, 0.4372, 0.1333, 0.1105, 0.0658, 0.0801], + device='cuda:0'), in_proj_covar=tensor([0.0628, 0.0646, 0.0695, 0.0627, 0.0708, 0.0603, 0.0606, 0.0677], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 04:08:58,258 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.1762, 4.1328, 3.7477, 1.9007, 3.6308, 3.7989, 3.7473, 3.6484], + device='cuda:0'), covar=tensor([0.0708, 0.0521, 0.0994, 0.4702, 0.0878, 0.0940, 0.1223, 0.0852], + device='cuda:0'), in_proj_covar=tensor([0.0522, 0.0433, 0.0432, 0.0536, 0.0424, 0.0444, 0.0424, 0.0387], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 04:09:26,990 INFO [optim.py:369] (0/4) 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 04:09:27,011 INFO [train.py:901] (0/4) Epoch 22, batch 5250, loss[loss=0.2268, simple_loss=0.3173, pruned_loss=0.06814, over 8337.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2884, pruned_loss=0.06207, over 1619039.05 frames. ], batch size: 26, lr: 3.43e-03, grad_scale: 8.0 +2023-02-07 04:09:31,811 WARNING [train.py:1067] (0/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-07 04:09:44,045 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175016.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:09:49,535 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175024.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:10:01,595 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175041.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:10:01,999 INFO [train.py:901] (0/4) Epoch 22, batch 5300, loss[loss=0.2476, simple_loss=0.307, pruned_loss=0.09413, over 7815.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.288, pruned_loss=0.06213, over 1611118.84 frames. ], batch size: 20, lr: 3.43e-03, grad_scale: 8.0 +2023-02-07 04:10:07,097 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175049.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:10:19,113 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=175067.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:10:35,756 INFO [optim.py:369] (0/4) 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 04:10:35,776 INFO [train.py:901] (0/4) Epoch 22, batch 5350, loss[loss=0.1824, simple_loss=0.2737, pruned_loss=0.04559, over 8192.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2865, pruned_loss=0.06145, over 1609008.31 frames. ], batch size: 23, lr: 3.43e-03, grad_scale: 8.0 +2023-02-07 04:10:57,891 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175121.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:11:11,825 INFO [train.py:901] (0/4) Epoch 22, batch 5400, loss[loss=0.2308, simple_loss=0.3167, pruned_loss=0.07247, over 8351.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2866, pruned_loss=0.06176, over 1608895.57 frames. ], batch size: 25, lr: 3.42e-03, grad_scale: 8.0 +2023-02-07 04:11:13,249 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.2305, 3.1841, 2.9085, 1.5595, 2.8017, 2.9360, 2.8538, 2.8798], + device='cuda:0'), covar=tensor([0.1178, 0.0799, 0.1283, 0.4588, 0.1104, 0.1326, 0.1523, 0.0987], + device='cuda:0'), in_proj_covar=tensor([0.0521, 0.0432, 0.0430, 0.0533, 0.0422, 0.0442, 0.0422, 0.0384], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 04:11:14,633 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175146.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:11:22,709 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175157.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:11:39,690 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=175182.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:11:46,276 INFO [optim.py:369] (0/4) 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 04:11:46,296 INFO [train.py:901] (0/4) Epoch 22, batch 5450, loss[loss=0.202, simple_loss=0.2856, pruned_loss=0.05923, over 8132.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2872, pruned_loss=0.06171, over 1613320.31 frames. ], batch size: 22, lr: 3.42e-03, grad_scale: 8.0 +2023-02-07 04:12:09,756 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175225.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:12:20,253 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-07 04:12:22,327 INFO [train.py:901] (0/4) Epoch 22, batch 5500, loss[loss=0.2329, simple_loss=0.3257, pruned_loss=0.07004, over 8527.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2864, pruned_loss=0.06148, over 1608499.38 frames. ], batch size: 28, lr: 3.42e-03, grad_scale: 8.0 +2023-02-07 04:12:40,055 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-02-07 04:12:56,619 INFO [optim.py:369] (0/4) 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 04:12:56,639 INFO [train.py:901] (0/4) Epoch 22, batch 5550, loss[loss=0.1916, simple_loss=0.2552, pruned_loss=0.06401, over 7419.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2865, pruned_loss=0.06126, over 1608236.14 frames. ], batch size: 17, lr: 3.42e-03, grad_scale: 8.0 +2023-02-07 04:13:31,790 INFO [train.py:901] (0/4) Epoch 22, batch 5600, loss[loss=0.1869, simple_loss=0.2637, pruned_loss=0.05505, over 7798.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2865, pruned_loss=0.0612, over 1614561.85 frames. ], batch size: 20, lr: 3.42e-03, grad_scale: 8.0 +2023-02-07 04:13:46,066 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6827, 2.0880, 3.2452, 1.4888, 2.4374, 2.1557, 1.6769, 2.5379], + device='cuda:0'), covar=tensor([0.1902, 0.2517, 0.0832, 0.4504, 0.1866, 0.3145, 0.2478, 0.2180], + device='cuda:0'), in_proj_covar=tensor([0.0525, 0.0604, 0.0554, 0.0641, 0.0645, 0.0589, 0.0535, 0.0629], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 04:13:49,337 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6417, 1.5952, 2.1181, 1.4211, 1.3108, 2.0874, 0.2697, 1.2418], + device='cuda:0'), covar=tensor([0.1663, 0.1521, 0.0359, 0.1039, 0.2546, 0.0428, 0.2122, 0.1419], + device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0197, 0.0129, 0.0220, 0.0267, 0.0137, 0.0168, 0.0194], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 04:13:50,743 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.16 vs. limit=5.0 +2023-02-07 04:14:03,191 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8416, 2.1310, 1.7477, 2.6613, 1.2902, 1.5884, 2.0044, 2.0888], + device='cuda:0'), covar=tensor([0.0762, 0.0746, 0.0923, 0.0366, 0.1014, 0.1274, 0.0674, 0.0698], + device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0198, 0.0247, 0.0216, 0.0208, 0.0247, 0.0252, 0.0209], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 04:14:03,816 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175388.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:14:06,408 INFO [optim.py:369] (0/4) 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 04:14:06,428 INFO [train.py:901] (0/4) Epoch 22, batch 5650, loss[loss=0.1918, simple_loss=0.2768, pruned_loss=0.05342, over 8284.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.287, pruned_loss=0.06165, over 1612462.23 frames. ], batch size: 23, lr: 3.42e-03, grad_scale: 8.0 +2023-02-07 04:14:09,950 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6815, 1.9489, 2.9872, 1.5023, 2.1801, 2.0111, 1.7025, 2.1098], + device='cuda:0'), covar=tensor([0.1796, 0.2384, 0.0825, 0.4345, 0.1839, 0.3202, 0.2278, 0.2210], + device='cuda:0'), in_proj_covar=tensor([0.0525, 0.0604, 0.0553, 0.0640, 0.0645, 0.0589, 0.0535, 0.0628], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 04:14:20,107 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175412.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:14:22,700 WARNING [train.py:1067] (0/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-07 04:14:37,560 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 22, batch 5700, loss[loss=0.1822, simple_loss=0.2608, pruned_loss=0.05181, over 7693.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2864, pruned_loss=0.06154, over 1613922.31 frames. ], batch size: 18, lr: 3.42e-03, grad_scale: 8.0 +2023-02-07 04:14:56,160 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175463.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:15:15,527 INFO [optim.py:369] (0/4) 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 04:15:15,547 INFO [train.py:901] (0/4) Epoch 22, batch 5750, loss[loss=0.2175, simple_loss=0.297, pruned_loss=0.06895, over 8485.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.287, pruned_loss=0.06172, over 1613430.64 frames. ], batch size: 25, lr: 3.42e-03, grad_scale: 8.0 +2023-02-07 04:15:20,411 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175499.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:15:22,435 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=175501.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:15:27,204 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-07 04:15:50,255 INFO [train.py:901] (0/4) Epoch 22, batch 5800, loss[loss=0.2118, simple_loss=0.293, pruned_loss=0.06527, over 8501.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2872, pruned_loss=0.06162, over 1613198.86 frames. ], batch size: 26, lr: 3.42e-03, grad_scale: 8.0 +2023-02-07 04:16:09,010 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=175569.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:16:25,901 INFO [optim.py:369] (0/4) 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 04:16:25,921 INFO [train.py:901] (0/4) Epoch 22, batch 5850, loss[loss=0.2008, simple_loss=0.28, pruned_loss=0.06077, over 7254.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2868, pruned_loss=0.06134, over 1613432.71 frames. ], batch size: 16, lr: 3.42e-03, grad_scale: 8.0 +2023-02-07 04:16:42,215 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=175616.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:17:00,355 INFO [train.py:901] (0/4) Epoch 22, batch 5900, loss[loss=0.2132, simple_loss=0.2916, pruned_loss=0.06737, over 8702.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2883, pruned_loss=0.06228, over 1617550.54 frames. ], batch size: 34, lr: 3.42e-03, grad_scale: 8.0 +2023-02-07 04:17:29,305 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) Epoch 22, batch 5950, loss[loss=0.1979, simple_loss=0.2686, pruned_loss=0.06355, over 7651.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2864, pruned_loss=0.0613, over 1614169.86 frames. ], batch size: 19, lr: 3.42e-03, grad_scale: 8.0 +2023-02-07 04:18:00,037 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175728.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:18:02,692 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=175732.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:18:09,579 INFO [train.py:901] (0/4) Epoch 22, batch 6000, loss[loss=0.185, simple_loss=0.2822, pruned_loss=0.0439, over 8507.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2872, pruned_loss=0.06149, over 1616366.72 frames. ], batch size: 28, lr: 3.42e-03, grad_scale: 8.0 +2023-02-07 04:18:09,579 INFO [train.py:926] (0/4) Computing validation loss +2023-02-07 04:18:21,639 INFO [train.py:935] (0/4) Epoch 22, validation: loss=0.1729, simple_loss=0.2732, pruned_loss=0.03632, over 944034.00 frames. +2023-02-07 04:18:21,640 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-07 04:18:31,458 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=175756.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:18:56,208 INFO [optim.py:369] (0/4) 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 04:18:56,229 INFO [train.py:901] (0/4) Epoch 22, batch 6050, loss[loss=0.2441, simple_loss=0.3184, pruned_loss=0.08487, over 8838.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2881, pruned_loss=0.06188, over 1616147.50 frames. ], batch size: 40, lr: 3.42e-03, grad_scale: 8.0 +2023-02-07 04:19:31,875 INFO [train.py:901] (0/4) Epoch 22, batch 6100, loss[loss=0.2056, simple_loss=0.302, pruned_loss=0.05454, over 8184.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2879, pruned_loss=0.06189, over 1615609.01 frames. ], batch size: 23, lr: 3.42e-03, grad_scale: 8.0 +2023-02-07 04:19:32,683 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=175843.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:19:35,630 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=175847.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:19:51,977 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=175871.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:19:52,699 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175872.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:19:56,549 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-07 04:20:02,529 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2018, 1.9456, 2.5209, 2.1096, 2.3786, 2.2171, 2.0359, 1.2230], + device='cuda:0'), covar=tensor([0.5051, 0.4470, 0.1925, 0.3539, 0.2475, 0.3040, 0.1804, 0.5223], + device='cuda:0'), in_proj_covar=tensor([0.0942, 0.0980, 0.0806, 0.0945, 0.0996, 0.0896, 0.0747, 0.0826], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 04:20:04,917 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 +2023-02-07 04:20:07,201 INFO [optim.py:369] (0/4) 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 04:20:07,221 INFO [train.py:901] (0/4) Epoch 22, batch 6150, loss[loss=0.1859, simple_loss=0.2714, pruned_loss=0.05022, over 8077.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2873, pruned_loss=0.06188, over 1610673.45 frames. ], batch size: 21, lr: 3.42e-03, grad_scale: 8.0 +2023-02-07 04:20:10,657 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175897.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:20:11,999 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175899.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:20:35,031 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.7556, 2.4084, 3.4123, 2.6695, 3.2177, 2.6557, 2.4585, 1.8742], + device='cuda:0'), covar=tensor([0.4962, 0.5009, 0.1699, 0.3505, 0.2478, 0.2840, 0.1755, 0.5428], + device='cuda:0'), in_proj_covar=tensor([0.0944, 0.0981, 0.0806, 0.0946, 0.0997, 0.0897, 0.0749, 0.0827], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 04:20:40,366 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175940.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:20:41,027 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0420, 2.4280, 2.7130, 1.3973, 2.8964, 1.6950, 1.5086, 2.1063], + device='cuda:0'), covar=tensor([0.0854, 0.0432, 0.0374, 0.0882, 0.0573, 0.0906, 0.0891, 0.0556], + device='cuda:0'), in_proj_covar=tensor([0.0454, 0.0390, 0.0346, 0.0447, 0.0378, 0.0533, 0.0391, 0.0421], + device='cuda:0'), out_proj_covar=tensor([1.2165e-04, 1.0223e-04, 9.0932e-05, 1.1754e-04, 9.9585e-05, 1.5018e-04, + 1.0525e-04, 1.1143e-04], device='cuda:0') +2023-02-07 04:20:41,476 INFO [train.py:901] (0/4) Epoch 22, batch 6200, loss[loss=0.1826, simple_loss=0.2789, pruned_loss=0.04314, over 8607.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.287, pruned_loss=0.06132, over 1611638.56 frames. ], batch size: 34, lr: 3.42e-03, grad_scale: 8.0 +2023-02-07 04:20:52,895 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=175958.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:20:56,912 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175963.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:20:58,221 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175965.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:21:15,640 INFO [optim.py:369] (0/4) 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 04:21:15,660 INFO [train.py:901] (0/4) Epoch 22, batch 6250, loss[loss=0.2072, simple_loss=0.2891, pruned_loss=0.06269, over 8642.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2875, pruned_loss=0.06206, over 1611355.14 frames. ], batch size: 39, lr: 3.42e-03, grad_scale: 16.0 +2023-02-07 04:21:21,790 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-176000.pt +2023-02-07 04:21:26,197 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.6680, 4.6600, 4.1909, 2.0737, 4.1454, 4.2073, 4.1282, 4.0691], + device='cuda:0'), covar=tensor([0.0614, 0.0480, 0.0996, 0.4465, 0.0741, 0.0872, 0.1261, 0.0693], + device='cuda:0'), in_proj_covar=tensor([0.0531, 0.0438, 0.0437, 0.0543, 0.0429, 0.0448, 0.0428, 0.0389], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 04:21:51,350 INFO [train.py:901] (0/4) Epoch 22, batch 6300, loss[loss=0.2039, simple_loss=0.289, pruned_loss=0.05945, over 8129.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2894, pruned_loss=0.06318, over 1613444.90 frames. ], batch size: 22, lr: 3.42e-03, grad_scale: 16.0 +2023-02-07 04:22:12,730 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=176072.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:22:26,685 INFO [optim.py:369] (0/4) 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 04:22:26,705 INFO [train.py:901] (0/4) Epoch 22, batch 6350, loss[loss=0.2527, simple_loss=0.3401, pruned_loss=0.08266, over 8722.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2874, pruned_loss=0.06188, over 1611501.82 frames. ], batch size: 30, lr: 3.42e-03, grad_scale: 16.0 +2023-02-07 04:22:27,806 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.13 vs. limit=5.0 +2023-02-07 04:22:29,245 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 +2023-02-07 04:22:34,393 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176103.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:22:51,184 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176127.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:22:51,828 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176128.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:23:01,397 INFO [train.py:901] (0/4) Epoch 22, batch 6400, loss[loss=0.1921, simple_loss=0.281, pruned_loss=0.05158, over 8374.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2873, pruned_loss=0.06163, over 1616506.28 frames. ], batch size: 24, lr: 3.41e-03, grad_scale: 16.0 +2023-02-07 04:23:08,431 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176152.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:23:33,469 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) Epoch 22, batch 6450, loss[loss=0.201, simple_loss=0.2903, pruned_loss=0.05587, over 8200.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2874, pruned_loss=0.06144, over 1617152.38 frames. ], batch size: 23, lr: 3.41e-03, grad_scale: 16.0 +2023-02-07 04:23:52,603 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176214.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:24:09,899 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176239.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:24:11,675 INFO [train.py:901] (0/4) Epoch 22, batch 6500, loss[loss=0.1794, simple_loss=0.2597, pruned_loss=0.04959, over 7685.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2876, pruned_loss=0.06125, over 1616265.06 frames. ], batch size: 18, lr: 3.41e-03, grad_scale: 16.0 +2023-02-07 04:24:12,441 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=176243.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:24:15,767 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1751, 1.5175, 4.4122, 2.0762, 2.6020, 5.0265, 5.1388, 4.3378], + device='cuda:0'), covar=tensor([0.1239, 0.2005, 0.0280, 0.1950, 0.1137, 0.0183, 0.0400, 0.0569], + device='cuda:0'), in_proj_covar=tensor([0.0295, 0.0323, 0.0287, 0.0315, 0.0310, 0.0266, 0.0421, 0.0306], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 04:24:23,871 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176260.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:24:34,064 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176275.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:24:41,794 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7183, 2.1186, 3.3106, 1.4782, 2.5424, 2.2058, 1.7792, 2.4876], + device='cuda:0'), covar=tensor([0.1830, 0.2702, 0.0811, 0.4647, 0.1741, 0.3025, 0.2333, 0.2131], + device='cuda:0'), in_proj_covar=tensor([0.0527, 0.0606, 0.0557, 0.0644, 0.0648, 0.0595, 0.0538, 0.0633], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 04:24:45,632 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 22, batch 6550, loss[loss=0.2298, simple_loss=0.3158, pruned_loss=0.0719, over 8808.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2873, pruned_loss=0.06093, over 1612154.51 frames. ], batch size: 39, lr: 3.41e-03, grad_scale: 16.0 +2023-02-07 04:24:57,260 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=176307.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:25:09,355 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-07 04:25:13,583 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3197, 2.1538, 1.8010, 2.0236, 1.8133, 1.4615, 1.7392, 1.6976], + device='cuda:0'), covar=tensor([0.1221, 0.0406, 0.1140, 0.0469, 0.0673, 0.1443, 0.0906, 0.0770], + device='cuda:0'), in_proj_covar=tensor([0.0352, 0.0232, 0.0333, 0.0309, 0.0298, 0.0340, 0.0345, 0.0316], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 04:25:21,087 INFO [train.py:901] (0/4) Epoch 22, batch 6600, loss[loss=0.1847, simple_loss=0.2525, pruned_loss=0.05849, over 7286.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2879, pruned_loss=0.06159, over 1611134.84 frames. ], batch size: 16, lr: 3.41e-03, grad_scale: 16.0 +2023-02-07 04:25:29,279 WARNING [train.py:1067] (0/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-07 04:25:32,750 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176358.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:25:55,382 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 22, batch 6650, loss[loss=0.1662, simple_loss=0.2448, pruned_loss=0.04384, over 7406.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2873, pruned_loss=0.06125, over 1615294.34 frames. ], batch size: 17, lr: 3.41e-03, grad_scale: 16.0 +2023-02-07 04:26:17,181 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176422.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:26:31,229 INFO [train.py:901] (0/4) Epoch 22, batch 6700, loss[loss=0.1894, simple_loss=0.2683, pruned_loss=0.05528, over 8241.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.288, pruned_loss=0.06179, over 1613395.61 frames. ], batch size: 22, lr: 3.41e-03, grad_scale: 16.0 +2023-02-07 04:26:32,076 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8864, 2.0423, 1.7218, 2.5608, 1.3364, 1.5528, 1.9629, 2.0299], + device='cuda:0'), covar=tensor([0.0730, 0.0708, 0.0886, 0.0362, 0.0996, 0.1244, 0.0713, 0.0750], + device='cuda:0'), in_proj_covar=tensor([0.0230, 0.0196, 0.0243, 0.0214, 0.0204, 0.0244, 0.0249, 0.0206], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 04:26:32,109 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176443.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:26:49,607 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176468.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:27:00,363 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176484.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:27:05,583 INFO [optim.py:369] (0/4) 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 04:27:05,602 INFO [train.py:901] (0/4) Epoch 22, batch 6750, loss[loss=0.2258, simple_loss=0.3106, pruned_loss=0.07053, over 8612.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2884, pruned_loss=0.06159, over 1616465.14 frames. ], batch size: 31, lr: 3.41e-03, grad_scale: 16.0 +2023-02-07 04:27:41,476 INFO [train.py:901] (0/4) Epoch 22, batch 6800, loss[loss=0.1991, simple_loss=0.2803, pruned_loss=0.05898, over 7252.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2886, pruned_loss=0.0624, over 1610553.43 frames. ], batch size: 16, lr: 3.41e-03, grad_scale: 16.0 +2023-02-07 04:27:44,989 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-07 04:28:16,785 INFO [optim.py:369] (0/4) 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 04:28:16,805 INFO [train.py:901] (0/4) Epoch 22, batch 6850, loss[loss=0.2339, simple_loss=0.2978, pruned_loss=0.08506, over 7808.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.288, pruned_loss=0.06221, over 1610209.43 frames. ], batch size: 20, lr: 3.41e-03, grad_scale: 16.0 +2023-02-07 04:28:24,733 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=176604.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:28:28,901 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6321, 2.0308, 3.0870, 1.3757, 2.4065, 1.8435, 1.8082, 2.2584], + device='cuda:0'), covar=tensor([0.2207, 0.2794, 0.1178, 0.5165, 0.2065, 0.3743, 0.2558, 0.2807], + device='cuda:0'), in_proj_covar=tensor([0.0526, 0.0605, 0.0555, 0.0643, 0.0648, 0.0594, 0.0537, 0.0632], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 04:28:31,632 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176614.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:28:34,701 WARNING [train.py:1067] (0/4) 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] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=176619.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:28:48,424 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176639.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:28:50,267 INFO [train.py:901] (0/4) Epoch 22, batch 6900, loss[loss=0.185, simple_loss=0.2691, pruned_loss=0.05044, over 7237.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2875, pruned_loss=0.0616, over 1610383.53 frames. ], batch size: 16, lr: 3.41e-03, grad_scale: 16.0 +2023-02-07 04:28:51,041 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176643.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:29:06,970 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.6374, 3.5787, 3.2994, 2.1769, 3.1851, 3.2339, 3.2682, 3.0477], + device='cuda:0'), covar=tensor([0.0856, 0.0659, 0.0999, 0.3753, 0.0954, 0.1122, 0.1247, 0.1088], + device='cuda:0'), in_proj_covar=tensor([0.0530, 0.0437, 0.0435, 0.0542, 0.0427, 0.0448, 0.0428, 0.0389], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 04:29:17,285 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176678.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:29:17,603 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 +2023-02-07 04:29:20,602 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176683.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:29:26,749 INFO [optim.py:369] (0/4) 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 04:29:26,775 INFO [train.py:901] (0/4) Epoch 22, batch 6950, loss[loss=0.1871, simple_loss=0.2783, pruned_loss=0.04789, over 8356.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2877, pruned_loss=0.06147, over 1611721.55 frames. ], batch size: 49, lr: 3.41e-03, grad_scale: 16.0 +2023-02-07 04:29:35,498 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176703.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:29:44,284 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-07 04:29:46,555 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176719.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:29:56,862 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176734.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:30:02,055 INFO [train.py:901] (0/4) Epoch 22, batch 7000, loss[loss=0.177, simple_loss=0.2698, pruned_loss=0.04207, over 8196.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2879, pruned_loss=0.06161, over 1612502.08 frames. ], batch size: 23, lr: 3.41e-03, grad_scale: 16.0 +2023-02-07 04:30:07,859 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 +2023-02-07 04:30:13,251 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.7204, 1.8477, 2.0203, 1.9071, 1.2013, 1.8057, 2.2856, 2.0008], + device='cuda:0'), covar=tensor([0.0465, 0.1106, 0.1498, 0.1225, 0.0600, 0.1316, 0.0623, 0.0598], + device='cuda:0'), in_proj_covar=tensor([0.0096, 0.0151, 0.0188, 0.0158, 0.0099, 0.0162, 0.0111, 0.0142], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-07 04:30:30,000 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6951, 1.8923, 1.9568, 1.4649, 2.0685, 1.4579, 0.5354, 1.9176], + device='cuda:0'), covar=tensor([0.0588, 0.0379, 0.0332, 0.0576, 0.0435, 0.0959, 0.0945, 0.0293], + device='cuda:0'), in_proj_covar=tensor([0.0453, 0.0390, 0.0345, 0.0447, 0.0379, 0.0535, 0.0391, 0.0422], + device='cuda:0'), out_proj_covar=tensor([1.2141e-04, 1.0229e-04, 9.0739e-05, 1.1767e-04, 9.9832e-05, 1.5080e-04, + 1.0535e-04, 1.1165e-04], device='cuda:0') +2023-02-07 04:30:37,819 INFO [optim.py:369] (0/4) 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 04:30:37,839 INFO [train.py:901] (0/4) Epoch 22, batch 7050, loss[loss=0.196, simple_loss=0.2832, pruned_loss=0.0544, over 8027.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2868, pruned_loss=0.06119, over 1608022.69 frames. ], batch size: 22, lr: 3.41e-03, grad_scale: 16.0 +2023-02-07 04:31:03,289 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=176828.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:31:12,380 INFO [train.py:901] (0/4) Epoch 22, batch 7100, loss[loss=0.202, simple_loss=0.2923, pruned_loss=0.05585, over 8026.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2868, pruned_loss=0.06074, over 1608551.48 frames. ], batch size: 22, lr: 3.41e-03, grad_scale: 16.0 +2023-02-07 04:31:30,665 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.49 vs. limit=2.0 +2023-02-07 04:31:31,805 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176871.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:31:46,098 INFO [optim.py:369] (0/4) 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 04:31:46,118 INFO [train.py:901] (0/4) Epoch 22, batch 7150, loss[loss=0.2422, simple_loss=0.3296, pruned_loss=0.07742, over 8501.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.287, pruned_loss=0.06065, over 1614071.54 frames. ], batch size: 49, lr: 3.41e-03, grad_scale: 16.0 +2023-02-07 04:32:22,266 INFO [train.py:901] (0/4) Epoch 22, batch 7200, loss[loss=0.1961, simple_loss=0.2711, pruned_loss=0.06061, over 6820.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2864, pruned_loss=0.06057, over 1607102.62 frames. ], batch size: 15, lr: 3.41e-03, grad_scale: 8.0 +2023-02-07 04:32:23,108 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176943.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:32:44,920 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176975.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:32:52,844 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=176987.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:32:54,993 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176990.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:32:56,132 INFO [train.py:901] (0/4) Epoch 22, batch 7250, loss[loss=0.2089, simple_loss=0.2776, pruned_loss=0.0701, over 7800.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.286, pruned_loss=0.06039, over 1609876.49 frames. ], batch size: 19, lr: 3.41e-03, grad_scale: 8.0 +2023-02-07 04:32:56,788 INFO [optim.py:369] (0/4) 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 04:33:02,397 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177000.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:33:14,103 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177015.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:33:21,887 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=177027.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:33:31,976 INFO [train.py:901] (0/4) Epoch 22, batch 7300, loss[loss=0.2119, simple_loss=0.2939, pruned_loss=0.06491, over 8129.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2864, pruned_loss=0.06031, over 1612517.97 frames. ], batch size: 22, lr: 3.41e-03, grad_scale: 8.0 +2023-02-07 04:34:06,491 INFO [train.py:901] (0/4) Epoch 22, batch 7350, loss[loss=0.1776, simple_loss=0.2521, pruned_loss=0.05151, over 7431.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2878, pruned_loss=0.0613, over 1610803.73 frames. ], batch size: 17, lr: 3.41e-03, grad_scale: 8.0 +2023-02-07 04:34:07,156 INFO [optim.py:369] (0/4) 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 04:34:13,478 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=177102.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:34:26,063 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-07 04:34:31,744 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=177127.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:34:33,086 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 22, batch 7400, loss[loss=0.2163, simple_loss=0.3056, pruned_loss=0.06352, over 8197.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.288, pruned_loss=0.06183, over 1612151.64 frames. ], batch size: 23, lr: 3.41e-03, grad_scale: 8.0 +2023-02-07 04:34:42,681 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=177142.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:34:47,994 WARNING [train.py:1067] (0/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-07 04:34:49,398 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3916, 1.6694, 1.6720, 0.9631, 1.7073, 1.3551, 0.2788, 1.6066], + device='cuda:0'), covar=tensor([0.0447, 0.0356, 0.0298, 0.0559, 0.0361, 0.0886, 0.0860, 0.0280], + device='cuda:0'), in_proj_covar=tensor([0.0458, 0.0396, 0.0348, 0.0451, 0.0383, 0.0540, 0.0394, 0.0425], + device='cuda:0'), out_proj_covar=tensor([1.2273e-04, 1.0389e-04, 9.1480e-05, 1.1879e-04, 1.0073e-04, 1.5222e-04, + 1.0607e-04, 1.1261e-04], device='cuda:0') +2023-02-07 04:35:16,507 INFO [train.py:901] (0/4) Epoch 22, batch 7450, loss[loss=0.1755, simple_loss=0.251, pruned_loss=0.05001, over 7189.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2883, pruned_loss=0.0623, over 1614439.60 frames. ], batch size: 16, lr: 3.40e-03, grad_scale: 8.0 +2023-02-07 04:35:17,191 INFO [optim.py:369] (0/4) 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 04:35:21,630 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177199.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:35:27,657 WARNING [train.py:1067] (0/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-07 04:35:32,251 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=177215.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:35:38,453 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 22, batch 7500, loss[loss=0.1882, simple_loss=0.2767, pruned_loss=0.04988, over 8026.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2876, pruned_loss=0.06183, over 1615052.63 frames. ], batch size: 22, lr: 3.40e-03, grad_scale: 8.0 +2023-02-07 04:36:13,912 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.76 vs. limit=5.0 +2023-02-07 04:36:15,019 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.1971, 4.1597, 3.7968, 1.9423, 3.7313, 3.7796, 3.7975, 3.6719], + device='cuda:0'), covar=tensor([0.0815, 0.0580, 0.1182, 0.4520, 0.0868, 0.0924, 0.1208, 0.0749], + device='cuda:0'), in_proj_covar=tensor([0.0531, 0.0441, 0.0434, 0.0541, 0.0427, 0.0449, 0.0429, 0.0390], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 04:36:25,359 INFO [train.py:901] (0/4) Epoch 22, batch 7550, loss[loss=0.215, simple_loss=0.3012, pruned_loss=0.06444, over 8358.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2873, pruned_loss=0.06138, over 1615820.91 frames. ], batch size: 24, lr: 3.40e-03, grad_scale: 8.0 +2023-02-07 04:36:26,046 INFO [optim.py:369] (0/4) 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 04:36:51,653 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=177330.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:36:59,620 INFO [train.py:901] (0/4) Epoch 22, batch 7600, loss[loss=0.2503, simple_loss=0.3278, pruned_loss=0.08644, over 8523.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2871, pruned_loss=0.06194, over 1608101.19 frames. ], batch size: 49, lr: 3.40e-03, grad_scale: 8.0 +2023-02-07 04:37:11,461 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177358.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:37:29,861 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177383.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:37:35,785 INFO [train.py:901] (0/4) Epoch 22, batch 7650, loss[loss=0.1681, simple_loss=0.2523, pruned_loss=0.04194, over 8086.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2869, pruned_loss=0.06155, over 1615292.41 frames. ], batch size: 21, lr: 3.40e-03, grad_scale: 8.0 +2023-02-07 04:37:36,441 INFO [optim.py:369] (0/4) 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 04:37:39,970 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177398.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:37:57,406 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177423.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:38:02,924 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8058, 1.9482, 2.0889, 1.4066, 2.2204, 1.5607, 0.7300, 1.8930], + device='cuda:0'), covar=tensor([0.0560, 0.0395, 0.0277, 0.0604, 0.0429, 0.0908, 0.0916, 0.0347], + device='cuda:0'), in_proj_covar=tensor([0.0453, 0.0393, 0.0344, 0.0447, 0.0379, 0.0535, 0.0391, 0.0422], + device='cuda:0'), out_proj_covar=tensor([1.2121e-04, 1.0304e-04, 9.0455e-05, 1.1757e-04, 9.9689e-05, 1.5079e-04, + 1.0516e-04, 1.1168e-04], device='cuda:0') +2023-02-07 04:38:09,398 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2643, 1.4694, 4.2661, 1.7353, 2.4710, 4.8007, 4.8923, 4.1634], + device='cuda:0'), covar=tensor([0.1201, 0.2128, 0.0281, 0.2266, 0.1221, 0.0194, 0.0486, 0.0582], + device='cuda:0'), in_proj_covar=tensor([0.0296, 0.0322, 0.0285, 0.0314, 0.0310, 0.0266, 0.0420, 0.0303], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 04:38:09,936 INFO [train.py:901] (0/4) Epoch 22, batch 7700, loss[loss=0.1862, simple_loss=0.2781, pruned_loss=0.04716, over 8487.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2877, pruned_loss=0.06207, over 1614726.03 frames. ], batch size: 29, lr: 3.40e-03, grad_scale: 8.0 +2023-02-07 04:38:14,876 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([0.9902, 1.6067, 1.4202, 1.5062, 1.3377, 1.2670, 1.2225, 1.2683], + device='cuda:0'), covar=tensor([0.1190, 0.0453, 0.1254, 0.0597, 0.0793, 0.1559, 0.0972, 0.0857], + device='cuda:0'), in_proj_covar=tensor([0.0355, 0.0232, 0.0335, 0.0309, 0.0299, 0.0340, 0.0344, 0.0315], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 04:38:30,444 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=177471.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:38:31,725 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=177473.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:38:38,630 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-07 04:38:45,979 INFO [train.py:901] (0/4) Epoch 22, batch 7750, loss[loss=0.2172, simple_loss=0.2991, pruned_loss=0.06765, over 8041.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2885, pruned_loss=0.06209, over 1614287.05 frames. ], batch size: 22, lr: 3.40e-03, grad_scale: 8.0 +2023-02-07 04:38:46,656 INFO [optim.py:369] (0/4) 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 04:38:49,023 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-07 04:39:20,414 INFO [train.py:901] (0/4) Epoch 22, batch 7800, loss[loss=0.2143, simple_loss=0.3047, pruned_loss=0.06195, over 8179.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2879, pruned_loss=0.06189, over 1611230.84 frames. ], batch size: 23, lr: 3.40e-03, grad_scale: 8.0 +2023-02-07 04:39:34,474 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1282, 1.9212, 2.4804, 2.1066, 2.4987, 2.2107, 2.0158, 1.2689], + device='cuda:0'), covar=tensor([0.5599, 0.4891, 0.1925, 0.3601, 0.2498, 0.3127, 0.1882, 0.5204], + device='cuda:0'), in_proj_covar=tensor([0.0946, 0.0981, 0.0811, 0.0950, 0.0999, 0.0899, 0.0754, 0.0829], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 04:39:39,765 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=177571.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:39:49,854 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=177586.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:39:49,890 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177586.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:39:51,147 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 22, batch 7850, loss[loss=0.1813, simple_loss=0.2574, pruned_loss=0.05259, over 7921.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2867, pruned_loss=0.06136, over 1612947.76 frames. ], batch size: 20, lr: 3.40e-03, grad_scale: 8.0 +2023-02-07 04:39:54,299 INFO [optim.py:369] (0/4) 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 04:40:06,552 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177611.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:40:26,685 INFO [train.py:901] (0/4) Epoch 22, batch 7900, loss[loss=0.2056, simple_loss=0.2981, pruned_loss=0.05654, over 8359.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2882, pruned_loss=0.06227, over 1609216.93 frames. ], batch size: 24, lr: 3.40e-03, grad_scale: 8.0 +2023-02-07 04:40:53,554 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 22, batch 7950, loss[loss=0.1854, simple_loss=0.2723, pruned_loss=0.04927, over 7918.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2885, pruned_loss=0.06272, over 1607749.29 frames. ], batch size: 20, lr: 3.40e-03, grad_scale: 8.0 +2023-02-07 04:41:00,683 INFO [optim.py:369] (0/4) 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 04:41:27,913 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9030, 1.3963, 3.4107, 1.4875, 2.3988, 3.7029, 3.7253, 3.2197], + device='cuda:0'), covar=tensor([0.1138, 0.1840, 0.0285, 0.1997, 0.0984, 0.0204, 0.0496, 0.0490], + device='cuda:0'), in_proj_covar=tensor([0.0298, 0.0323, 0.0286, 0.0316, 0.0312, 0.0267, 0.0423, 0.0305], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 04:41:33,720 INFO [train.py:901] (0/4) Epoch 22, batch 8000, loss[loss=0.2448, simple_loss=0.3262, pruned_loss=0.08171, over 8287.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2888, pruned_loss=0.06278, over 1611162.87 frames. ], batch size: 23, lr: 3.40e-03, grad_scale: 8.0 +2023-02-07 04:41:35,162 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.7317, 4.6943, 4.3000, 2.1823, 4.2239, 4.4049, 4.2936, 4.2296], + device='cuda:0'), covar=tensor([0.0607, 0.0438, 0.0916, 0.4454, 0.0719, 0.0748, 0.1118, 0.0666], + device='cuda:0'), in_proj_covar=tensor([0.0524, 0.0434, 0.0429, 0.0532, 0.0420, 0.0441, 0.0420, 0.0385], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 04:41:44,709 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4671, 1.6492, 2.1722, 1.3377, 1.4101, 1.6624, 1.5938, 1.4935], + device='cuda:0'), covar=tensor([0.2224, 0.2861, 0.1101, 0.4901, 0.2300, 0.3805, 0.2589, 0.2280], + device='cuda:0'), in_proj_covar=tensor([0.0528, 0.0607, 0.0557, 0.0649, 0.0650, 0.0596, 0.0538, 0.0633], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 04:42:06,689 INFO [train.py:901] (0/4) Epoch 22, batch 8050, loss[loss=0.1677, simple_loss=0.2467, pruned_loss=0.04432, over 7545.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2876, pruned_loss=0.06282, over 1596005.00 frames. ], batch size: 18, lr: 3.40e-03, grad_scale: 8.0 +2023-02-07 04:42:07,275 INFO [optim.py:369] (0/4) 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 04:42:10,775 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=177798.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:42:22,339 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=177815.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:42:25,487 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.58 vs. limit=5.0 +2023-02-07 04:42:30,249 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/epoch-22.pt +2023-02-07 04:42:41,458 WARNING [train.py:1067] (0/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-07 04:42:44,818 INFO [train.py:901] (0/4) Epoch 23, batch 0, loss[loss=0.1775, simple_loss=0.2533, pruned_loss=0.05085, over 7239.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2533, pruned_loss=0.05085, over 7239.00 frames. ], batch size: 16, lr: 3.32e-03, grad_scale: 8.0 +2023-02-07 04:42:44,818 INFO [train.py:926] (0/4) Computing validation loss +2023-02-07 04:42:56,156 INFO [train.py:935] (0/4) Epoch 23, validation: loss=0.1743, simple_loss=0.274, pruned_loss=0.0373, over 944034.00 frames. +2023-02-07 04:42:56,157 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-07 04:43:08,351 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177842.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:43:10,527 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177844.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:43:12,415 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-07 04:43:20,505 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.6931, 1.3712, 5.8324, 2.0316, 5.2297, 4.8804, 5.4140, 5.2486], + device='cuda:0'), covar=tensor([0.0535, 0.5455, 0.0352, 0.4341, 0.1011, 0.0917, 0.0485, 0.0543], + device='cuda:0'), in_proj_covar=tensor([0.0634, 0.0648, 0.0700, 0.0632, 0.0709, 0.0606, 0.0610, 0.0684], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 04:43:25,876 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177867.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:43:28,075 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177869.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:43:32,006 INFO [train.py:901] (0/4) Epoch 23, batch 50, loss[loss=0.1896, simple_loss=0.2802, pruned_loss=0.04947, over 8486.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2904, pruned_loss=0.06287, over 363679.24 frames. ], batch size: 49, lr: 3.32e-03, grad_scale: 8.0 +2023-02-07 04:43:45,269 INFO [optim.py:369] (0/4) 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] (0/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-07 04:44:01,098 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=177915.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:44:06,090 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0557, 1.6388, 1.3748, 1.5041, 1.3237, 1.3047, 1.2794, 1.2983], + device='cuda:0'), covar=tensor([0.1121, 0.0442, 0.1325, 0.0573, 0.0754, 0.1428, 0.0935, 0.0822], + device='cuda:0'), in_proj_covar=tensor([0.0351, 0.0230, 0.0330, 0.0307, 0.0296, 0.0336, 0.0341, 0.0312], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 04:44:07,964 INFO [train.py:901] (0/4) Epoch 23, batch 100, loss[loss=0.2125, simple_loss=0.3012, pruned_loss=0.06187, over 8473.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2872, pruned_loss=0.06064, over 644533.57 frames. ], batch size: 27, lr: 3.32e-03, grad_scale: 8.0 +2023-02-07 04:44:09,369 WARNING [train.py:1067] (0/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-07 04:44:42,228 INFO [train.py:901] (0/4) Epoch 23, batch 150, loss[loss=0.1824, simple_loss=0.2867, pruned_loss=0.03908, over 8461.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2893, pruned_loss=0.06157, over 863038.20 frames. ], batch size: 25, lr: 3.32e-03, grad_scale: 8.0 +2023-02-07 04:44:54,770 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 +2023-02-07 04:44:54,924 INFO [optim.py:369] (0/4) 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 04:44:59,737 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-178000.pt +2023-02-07 04:45:17,376 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 +2023-02-07 04:45:18,316 INFO [train.py:901] (0/4) Epoch 23, batch 200, loss[loss=0.1804, simple_loss=0.2605, pruned_loss=0.05021, over 7420.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2914, pruned_loss=0.06238, over 1035072.39 frames. ], batch size: 17, lr: 3.32e-03, grad_scale: 8.0 +2023-02-07 04:45:19,125 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=178026.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:45:21,848 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 23, batch 250, loss[loss=0.1891, simple_loss=0.262, pruned_loss=0.05804, over 7818.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2896, pruned_loss=0.06198, over 1163574.30 frames. ], batch size: 20, lr: 3.32e-03, grad_scale: 8.0 +2023-02-07 04:46:04,787 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-07 04:46:06,105 INFO [optim.py:369] (0/4) 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 04:46:06,309 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4847, 1.2973, 2.2664, 1.2395, 2.1606, 2.4395, 2.5519, 2.0727], + device='cuda:0'), covar=tensor([0.1009, 0.1328, 0.0459, 0.1908, 0.0755, 0.0362, 0.0757, 0.0671], + device='cuda:0'), in_proj_covar=tensor([0.0297, 0.0320, 0.0286, 0.0315, 0.0311, 0.0266, 0.0422, 0.0305], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 04:46:12,825 WARNING [train.py:1067] (0/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-07 04:46:23,219 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4705, 1.7443, 2.6351, 1.3575, 1.9439, 1.8389, 1.5274, 2.0006], + device='cuda:0'), covar=tensor([0.2097, 0.2735, 0.0961, 0.4686, 0.1898, 0.3310, 0.2443, 0.2200], + device='cuda:0'), in_proj_covar=tensor([0.0525, 0.0603, 0.0555, 0.0644, 0.0647, 0.0590, 0.0535, 0.0631], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 04:46:28,466 INFO [train.py:901] (0/4) Epoch 23, batch 300, loss[loss=0.1997, simple_loss=0.2856, pruned_loss=0.05683, over 8593.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2903, pruned_loss=0.06262, over 1266535.47 frames. ], batch size: 49, lr: 3.32e-03, grad_scale: 8.0 +2023-02-07 04:46:40,061 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=178141.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:46:40,621 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=178142.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:46:52,875 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=178159.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:46:55,262 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 +2023-02-07 04:47:03,748 INFO [train.py:901] (0/4) Epoch 23, batch 350, loss[loss=0.1888, simple_loss=0.2624, pruned_loss=0.05758, over 7773.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.289, pruned_loss=0.06195, over 1342021.34 frames. ], batch size: 19, lr: 3.32e-03, grad_scale: 8.0 +2023-02-07 04:47:16,034 INFO [optim.py:369] (0/4) 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 04:47:17,665 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4218, 2.0379, 3.0446, 1.7546, 1.5783, 2.9722, 0.9379, 2.0366], + device='cuda:0'), covar=tensor([0.1839, 0.1234, 0.0275, 0.1507, 0.2657, 0.0361, 0.2161, 0.1399], + device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0201, 0.0130, 0.0224, 0.0272, 0.0139, 0.0173, 0.0197], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 04:47:38,674 INFO [train.py:901] (0/4) Epoch 23, batch 400, loss[loss=0.2547, simple_loss=0.3428, pruned_loss=0.08329, over 8731.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2886, pruned_loss=0.06165, over 1404155.61 frames. ], batch size: 30, lr: 3.32e-03, grad_scale: 8.0 +2023-02-07 04:48:02,274 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=178257.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:48:02,315 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0843, 1.2298, 1.2337, 0.8126, 1.2338, 1.0290, 0.0839, 1.1970], + device='cuda:0'), covar=tensor([0.0419, 0.0415, 0.0367, 0.0544, 0.0421, 0.1077, 0.0892, 0.0356], + device='cuda:0'), in_proj_covar=tensor([0.0452, 0.0393, 0.0344, 0.0445, 0.0378, 0.0535, 0.0392, 0.0423], + device='cuda:0'), out_proj_covar=tensor([1.2080e-04, 1.0296e-04, 9.0392e-05, 1.1706e-04, 9.9405e-05, 1.5071e-04, + 1.0550e-04, 1.1208e-04], device='cuda:0') +2023-02-07 04:48:15,030 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=178274.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:48:15,485 INFO [train.py:901] (0/4) Epoch 23, batch 450, loss[loss=0.1984, simple_loss=0.2779, pruned_loss=0.05949, over 8244.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2881, pruned_loss=0.0617, over 1452029.80 frames. ], batch size: 24, lr: 3.32e-03, grad_scale: 8.0 +2023-02-07 04:48:20,714 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-07 04:48:23,152 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178286.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:48:27,632 INFO [optim.py:369] (0/4) 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 04:48:40,224 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=178311.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:48:50,176 INFO [train.py:901] (0/4) Epoch 23, batch 500, loss[loss=0.2154, simple_loss=0.2855, pruned_loss=0.07269, over 7822.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2885, pruned_loss=0.06169, over 1492497.95 frames. ], batch size: 20, lr: 3.32e-03, grad_scale: 8.0 +2023-02-07 04:48:50,378 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.0644, 1.7558, 1.9911, 1.6294, 1.0531, 1.7148, 2.2553, 2.1445], + device='cuda:0'), covar=tensor([0.0411, 0.1273, 0.1611, 0.1378, 0.0613, 0.1403, 0.0602, 0.0598], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0153, 0.0190, 0.0159, 0.0100, 0.0163, 0.0112, 0.0143], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-07 04:49:16,606 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 +2023-02-07 04:49:25,958 INFO [train.py:901] (0/4) Epoch 23, batch 550, loss[loss=0.2535, simple_loss=0.3292, pruned_loss=0.08895, over 8551.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2894, pruned_loss=0.0624, over 1517011.59 frames. ], batch size: 49, lr: 3.32e-03, grad_scale: 8.0 +2023-02-07 04:49:39,365 INFO [optim.py:369] (0/4) 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 04:49:42,424 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178397.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:49:59,307 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 23, batch 600, loss[loss=0.2027, simple_loss=0.2957, pruned_loss=0.05483, over 8328.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2885, pruned_loss=0.06144, over 1542126.90 frames. ], batch size: 25, lr: 3.32e-03, grad_scale: 8.0 +2023-02-07 04:50:14,790 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-07 04:50:33,524 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=178470.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:50:36,800 INFO [train.py:901] (0/4) Epoch 23, batch 650, loss[loss=0.1965, simple_loss=0.285, pruned_loss=0.05403, over 8099.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.287, pruned_loss=0.06101, over 1555435.05 frames. ], batch size: 23, lr: 3.32e-03, grad_scale: 8.0 +2023-02-07 04:50:49,810 INFO [optim.py:369] (0/4) 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 04:50:55,492 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5576, 1.5192, 2.1360, 1.3838, 1.2035, 2.0725, 0.4334, 1.2848], + device='cuda:0'), covar=tensor([0.1944, 0.1262, 0.0330, 0.0982, 0.2638, 0.0404, 0.2007, 0.1391], + device='cuda:0'), in_proj_covar=tensor([0.0194, 0.0202, 0.0131, 0.0224, 0.0272, 0.0139, 0.0173, 0.0198], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 04:51:04,340 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178513.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:51:11,890 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3465, 2.6603, 3.0307, 1.6607, 3.2990, 2.0769, 1.6783, 2.2921], + device='cuda:0'), covar=tensor([0.0913, 0.0426, 0.0318, 0.0824, 0.0504, 0.0832, 0.0963, 0.0538], + device='cuda:0'), in_proj_covar=tensor([0.0449, 0.0390, 0.0343, 0.0443, 0.0377, 0.0531, 0.0389, 0.0419], + device='cuda:0'), out_proj_covar=tensor([1.2001e-04, 1.0232e-04, 9.0183e-05, 1.1660e-04, 9.9026e-05, 1.4976e-04, + 1.0476e-04, 1.1106e-04], device='cuda:0') +2023-02-07 04:51:12,421 INFO [train.py:901] (0/4) Epoch 23, batch 700, loss[loss=0.1644, simple_loss=0.2525, pruned_loss=0.03812, over 7655.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2865, pruned_loss=0.06077, over 1569784.39 frames. ], batch size: 19, lr: 3.32e-03, grad_scale: 8.0 +2023-02-07 04:51:16,059 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178530.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:51:21,517 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=178538.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:51:32,923 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 +2023-02-07 04:51:33,982 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 23, batch 750, loss[loss=0.1824, simple_loss=0.2509, pruned_loss=0.05698, over 7245.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2877, pruned_loss=0.06156, over 1581166.51 frames. ], batch size: 16, lr: 3.32e-03, grad_scale: 8.0 +2023-02-07 04:52:00,640 INFO [optim.py:369] (0/4) 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 04:52:03,330 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-07 04:52:12,914 WARNING [train.py:1067] (0/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-07 04:52:13,852 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5617, 2.4953, 1.7913, 2.3984, 2.1965, 1.5011, 2.0990, 2.2138], + device='cuda:0'), covar=tensor([0.1619, 0.0453, 0.1354, 0.0604, 0.0745, 0.1705, 0.1053, 0.1014], + device='cuda:0'), in_proj_covar=tensor([0.0356, 0.0235, 0.0337, 0.0311, 0.0300, 0.0342, 0.0347, 0.0318], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 04:52:24,031 INFO [train.py:901] (0/4) Epoch 23, batch 800, loss[loss=0.1717, simple_loss=0.2582, pruned_loss=0.04266, over 8187.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2865, pruned_loss=0.06069, over 1584886.72 frames. ], batch size: 23, lr: 3.32e-03, grad_scale: 8.0 +2023-02-07 04:52:32,108 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=178637.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:52:57,754 INFO [train.py:901] (0/4) Epoch 23, batch 850, loss[loss=0.2219, simple_loss=0.2996, pruned_loss=0.07209, over 8023.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2877, pruned_loss=0.06147, over 1592220.15 frames. ], batch size: 22, lr: 3.31e-03, grad_scale: 8.0 +2023-02-07 04:53:10,567 INFO [optim.py:369] (0/4) 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 04:53:24,463 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=178712.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:53:24,481 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9840, 1.7378, 3.1929, 1.5543, 2.4367, 3.4286, 3.5456, 2.9560], + device='cuda:0'), covar=tensor([0.1097, 0.1545, 0.0334, 0.1897, 0.0891, 0.0254, 0.0667, 0.0557], + device='cuda:0'), in_proj_covar=tensor([0.0296, 0.0322, 0.0287, 0.0315, 0.0311, 0.0268, 0.0423, 0.0305], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 04:53:26,475 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=178715.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 04:53:27,219 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.1652, 1.0736, 1.3033, 0.9900, 0.9560, 1.3215, 0.0515, 0.8827], + device='cuda:0'), covar=tensor([0.1699, 0.1274, 0.0475, 0.0915, 0.2473, 0.0561, 0.2058, 0.1251], + device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0201, 0.0130, 0.0221, 0.0270, 0.0138, 0.0170, 0.0196], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 04:53:34,044 INFO [train.py:901] (0/4) Epoch 23, batch 900, loss[loss=0.2274, simple_loss=0.3106, pruned_loss=0.07212, over 7056.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2869, pruned_loss=0.06095, over 1595714.00 frames. ], batch size: 72, lr: 3.31e-03, grad_scale: 8.0 +2023-02-07 04:53:57,336 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.62 vs. limit=2.0 +2023-02-07 04:54:09,419 INFO [train.py:901] (0/4) Epoch 23, batch 950, loss[loss=0.17, simple_loss=0.2472, pruned_loss=0.04643, over 7939.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2867, pruned_loss=0.06087, over 1606081.82 frames. ], batch size: 20, lr: 3.31e-03, grad_scale: 8.0 +2023-02-07 04:54:18,536 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=178788.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:54:18,783 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.72 vs. limit=5.0 +2023-02-07 04:54:21,857 INFO [optim.py:369] (0/4) 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 04:54:35,811 WARNING [train.py:1067] (0/4) 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] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=178814.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:54:42,178 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5736, 2.0844, 3.2580, 1.4928, 2.4340, 2.0172, 1.7713, 2.4375], + device='cuda:0'), covar=tensor([0.1922, 0.2466, 0.0732, 0.4430, 0.1744, 0.3103, 0.2276, 0.2071], + device='cuda:0'), in_proj_covar=tensor([0.0521, 0.0598, 0.0549, 0.0639, 0.0641, 0.0586, 0.0530, 0.0626], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 04:54:45,361 INFO [train.py:901] (0/4) Epoch 23, batch 1000, loss[loss=0.1814, simple_loss=0.2661, pruned_loss=0.04839, over 8086.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2869, pruned_loss=0.06094, over 1607255.16 frames. ], batch size: 21, lr: 3.31e-03, grad_scale: 8.0 +2023-02-07 04:55:12,377 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-07 04:55:21,365 INFO [train.py:901] (0/4) Epoch 23, batch 1050, loss[loss=0.228, simple_loss=0.3105, pruned_loss=0.0727, over 8556.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.286, pruned_loss=0.06019, over 1613138.50 frames. ], batch size: 31, lr: 3.31e-03, grad_scale: 8.0 +2023-02-07 04:55:25,394 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-07 04:55:33,402 INFO [optim.py:369] (0/4) 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 04:55:46,647 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=178912.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 04:55:47,284 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1177, 1.7020, 4.5485, 2.0169, 2.6573, 5.1704, 5.2596, 4.5273], + device='cuda:0'), covar=tensor([0.1306, 0.1814, 0.0267, 0.1867, 0.1055, 0.0162, 0.0403, 0.0537], + device='cuda:0'), in_proj_covar=tensor([0.0296, 0.0320, 0.0287, 0.0315, 0.0310, 0.0267, 0.0422, 0.0305], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 04:55:56,202 INFO [train.py:901] (0/4) Epoch 23, batch 1100, loss[loss=0.214, simple_loss=0.2973, pruned_loss=0.06534, over 8444.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2863, pruned_loss=0.06095, over 1609796.92 frames. ], batch size: 27, lr: 3.31e-03, grad_scale: 16.0 +2023-02-07 04:55:59,143 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=178929.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:56:32,128 INFO [train.py:901] (0/4) Epoch 23, batch 1150, loss[loss=0.174, simple_loss=0.2596, pruned_loss=0.04417, over 7818.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2856, pruned_loss=0.06091, over 1605748.46 frames. ], batch size: 20, lr: 3.31e-03, grad_scale: 16.0 +2023-02-07 04:56:36,255 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-07 04:56:36,321 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=178981.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:56:45,237 INFO [optim.py:369] (0/4) 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 04:57:07,126 INFO [train.py:901] (0/4) Epoch 23, batch 1200, loss[loss=0.179, simple_loss=0.267, pruned_loss=0.04546, over 7787.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2856, pruned_loss=0.06088, over 1604739.51 frames. ], batch size: 19, lr: 3.31e-03, grad_scale: 16.0 +2023-02-07 04:57:29,067 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=179056.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:57:31,022 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=179059.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 04:57:42,802 INFO [train.py:901] (0/4) Epoch 23, batch 1250, loss[loss=0.2089, simple_loss=0.2833, pruned_loss=0.06722, over 8137.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2856, pruned_loss=0.06093, over 1608896.51 frames. ], batch size: 22, lr: 3.31e-03, grad_scale: 16.0 +2023-02-07 04:57:55,996 INFO [optim.py:369] (0/4) 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 04:57:58,278 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179096.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:58:05,898 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.1243, 1.0333, 1.2126, 0.9572, 0.9423, 1.2360, 0.1084, 0.9043], + device='cuda:0'), covar=tensor([0.1606, 0.1313, 0.0526, 0.0820, 0.2269, 0.0528, 0.2000, 0.1291], + device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0200, 0.0129, 0.0220, 0.0268, 0.0136, 0.0170, 0.0194], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 04:58:19,011 INFO [train.py:901] (0/4) Epoch 23, batch 1300, loss[loss=0.2074, simple_loss=0.2979, pruned_loss=0.05839, over 8024.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2868, pruned_loss=0.06162, over 1611789.27 frames. ], batch size: 22, lr: 3.31e-03, grad_scale: 16.0 +2023-02-07 04:58:20,698 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3801, 2.3138, 1.6993, 2.1240, 1.9716, 1.4821, 1.9516, 1.8516], + device='cuda:0'), covar=tensor([0.1454, 0.0415, 0.1297, 0.0556, 0.0771, 0.1534, 0.0973, 0.0966], + device='cuda:0'), in_proj_covar=tensor([0.0353, 0.0232, 0.0334, 0.0308, 0.0298, 0.0338, 0.0342, 0.0314], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 04:58:24,100 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=179132.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:58:51,876 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179171.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:58:53,999 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179174.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 04:58:54,478 INFO [train.py:901] (0/4) Epoch 23, batch 1350, loss[loss=0.1903, simple_loss=0.2751, pruned_loss=0.05279, over 8241.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2859, pruned_loss=0.06109, over 1611662.07 frames. ], batch size: 22, lr: 3.31e-03, grad_scale: 16.0 +2023-02-07 04:58:55,432 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3516, 2.2550, 2.9887, 2.4541, 2.8962, 2.4737, 2.2253, 1.7847], + device='cuda:0'), covar=tensor([0.5521, 0.5387, 0.2026, 0.3696, 0.2615, 0.2943, 0.1870, 0.5493], + device='cuda:0'), in_proj_covar=tensor([0.0938, 0.0984, 0.0811, 0.0950, 0.0994, 0.0898, 0.0752, 0.0828], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 04:59:01,685 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 04:59:16,580 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 +2023-02-07 04:59:20,414 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179210.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:59:30,635 INFO [train.py:901] (0/4) Epoch 23, batch 1400, loss[loss=0.1915, simple_loss=0.2881, pruned_loss=0.04742, over 8241.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2861, pruned_loss=0.06095, over 1611512.27 frames. ], batch size: 24, lr: 3.31e-03, grad_scale: 16.0 +2023-02-07 04:59:37,572 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 +2023-02-07 04:59:47,155 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179247.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 04:59:53,424 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=179256.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 04:59:57,850 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 +2023-02-07 05:00:06,550 INFO [train.py:901] (0/4) Epoch 23, batch 1450, loss[loss=0.1663, simple_loss=0.2552, pruned_loss=0.03874, over 7673.00 frames. ], tot_loss[loss=0.205, simple_loss=0.287, pruned_loss=0.06152, over 1612883.88 frames. ], batch size: 19, lr: 3.31e-03, grad_scale: 16.0 +2023-02-07 05:00:16,906 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-07 05:00:19,769 INFO [optim.py:369] (0/4) 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 05:00:43,616 INFO [train.py:901] (0/4) Epoch 23, batch 1500, loss[loss=0.2028, simple_loss=0.2877, pruned_loss=0.05892, over 8248.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2873, pruned_loss=0.06198, over 1612111.75 frames. ], batch size: 22, lr: 3.31e-03, grad_scale: 16.0 +2023-02-07 05:00:57,219 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7934, 1.8265, 2.5019, 1.5862, 1.3949, 2.4545, 0.4965, 1.4932], + device='cuda:0'), covar=tensor([0.1907, 0.1179, 0.0287, 0.1304, 0.2569, 0.0326, 0.1993, 0.1377], + device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0199, 0.0129, 0.0219, 0.0267, 0.0136, 0.0169, 0.0193], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 05:01:03,389 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179352.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:01:16,374 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179371.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 05:01:18,855 INFO [train.py:901] (0/4) Epoch 23, batch 1550, loss[loss=0.1714, simple_loss=0.25, pruned_loss=0.0464, over 7775.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2866, pruned_loss=0.06158, over 1609471.78 frames. ], batch size: 19, lr: 3.31e-03, grad_scale: 16.0 +2023-02-07 05:01:20,456 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179379.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:01:31,094 INFO [optim.py:369] (0/4) 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 05:01:32,066 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2440, 2.1748, 1.7155, 1.9682, 1.7870, 1.5079, 1.6689, 1.7150], + device='cuda:0'), covar=tensor([0.1341, 0.0375, 0.1193, 0.0487, 0.0746, 0.1421, 0.0926, 0.0849], + device='cuda:0'), in_proj_covar=tensor([0.0354, 0.0234, 0.0336, 0.0311, 0.0300, 0.0339, 0.0346, 0.0317], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 05:01:54,027 INFO [train.py:901] (0/4) Epoch 23, batch 1600, loss[loss=0.2088, simple_loss=0.2929, pruned_loss=0.06235, over 8237.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2872, pruned_loss=0.06169, over 1612129.12 frames. ], batch size: 22, lr: 3.31e-03, grad_scale: 16.0 +2023-02-07 05:01:56,459 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179427.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:01:58,497 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179430.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 05:02:14,544 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179452.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:02:16,562 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179455.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 05:02:21,374 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179462.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 05:02:31,190 INFO [train.py:901] (0/4) Epoch 23, batch 1650, loss[loss=0.2024, simple_loss=0.2808, pruned_loss=0.06197, over 7534.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.287, pruned_loss=0.06115, over 1616263.29 frames. ], batch size: 18, lr: 3.31e-03, grad_scale: 16.0 +2023-02-07 05:02:43,578 INFO [optim.py:369] (0/4) 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 05:02:50,690 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 23, batch 1700, loss[loss=0.2133, simple_loss=0.2879, pruned_loss=0.06936, over 7783.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2864, pruned_loss=0.06067, over 1614851.69 frames. ], batch size: 19, lr: 3.31e-03, grad_scale: 16.0 +2023-02-07 05:03:08,681 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179528.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:03:32,848 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5542, 1.9276, 3.0007, 1.4060, 2.2613, 1.9285, 1.6852, 2.3045], + device='cuda:0'), covar=tensor([0.1966, 0.2541, 0.0842, 0.4585, 0.1840, 0.3265, 0.2304, 0.2117], + device='cuda:0'), in_proj_covar=tensor([0.0527, 0.0608, 0.0556, 0.0647, 0.0653, 0.0595, 0.0538, 0.0633], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 05:03:42,338 INFO [train.py:901] (0/4) Epoch 23, batch 1750, loss[loss=0.1677, simple_loss=0.2426, pruned_loss=0.04646, over 7209.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2864, pruned_loss=0.06043, over 1614642.53 frames. ], batch size: 16, lr: 3.31e-03, grad_scale: 8.0 +2023-02-07 05:03:44,861 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 +2023-02-07 05:03:52,465 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 +2023-02-07 05:03:56,235 INFO [optim.py:369] (0/4) 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 05:03:58,697 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6453, 2.0306, 2.1834, 1.2807, 2.2408, 1.5863, 0.7598, 1.9037], + device='cuda:0'), covar=tensor([0.0666, 0.0395, 0.0273, 0.0674, 0.0467, 0.0890, 0.0897, 0.0394], + device='cuda:0'), in_proj_covar=tensor([0.0454, 0.0394, 0.0348, 0.0449, 0.0382, 0.0535, 0.0394, 0.0422], + device='cuda:0'), out_proj_covar=tensor([1.2119e-04, 1.0329e-04, 9.1232e-05, 1.1818e-04, 1.0075e-04, 1.5078e-04, + 1.0621e-04, 1.1164e-04], device='cuda:0') +2023-02-07 05:04:15,366 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.3702, 1.6658, 4.5542, 1.7287, 4.0218, 3.7705, 4.1689, 3.9912], + device='cuda:0'), covar=tensor([0.0614, 0.4665, 0.0516, 0.4293, 0.1201, 0.0958, 0.0573, 0.0704], + device='cuda:0'), in_proj_covar=tensor([0.0640, 0.0651, 0.0707, 0.0640, 0.0717, 0.0613, 0.0612, 0.0690], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 05:04:17,968 INFO [train.py:901] (0/4) Epoch 23, batch 1800, loss[loss=0.2236, simple_loss=0.2918, pruned_loss=0.07773, over 8666.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2857, pruned_loss=0.06011, over 1618861.44 frames. ], batch size: 34, lr: 3.31e-03, grad_scale: 8.0 +2023-02-07 05:04:19,554 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179627.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 05:04:37,264 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179652.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 05:04:38,538 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 23, batch 1850, loss[loss=0.1881, simple_loss=0.2745, pruned_loss=0.05083, over 8141.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.286, pruned_loss=0.06047, over 1618375.64 frames. ], batch size: 22, lr: 3.31e-03, grad_scale: 8.0 +2023-02-07 05:05:00,031 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1478, 1.5911, 1.9636, 1.5226, 1.0353, 1.6586, 1.9230, 1.8478], + device='cuda:0'), covar=tensor([0.0512, 0.1230, 0.1569, 0.1356, 0.0574, 0.1362, 0.0618, 0.0604], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0153, 0.0190, 0.0159, 0.0100, 0.0163, 0.0112, 0.0144], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-07 05:05:07,477 INFO [optim.py:369] (0/4) 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 05:05:28,483 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=179723.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:05:29,811 INFO [train.py:901] (0/4) Epoch 23, batch 1900, loss[loss=0.2271, simple_loss=0.3126, pruned_loss=0.07084, over 8576.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2864, pruned_loss=0.06032, over 1618513.20 frames. ], batch size: 31, lr: 3.31e-03, grad_scale: 8.0 +2023-02-07 05:05:59,884 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-07 05:06:05,567 INFO [train.py:901] (0/4) Epoch 23, batch 1950, loss[loss=0.1694, simple_loss=0.2538, pruned_loss=0.04247, over 8143.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2869, pruned_loss=0.06056, over 1616319.29 frames. ], batch size: 22, lr: 3.30e-03, grad_scale: 8.0 +2023-02-07 05:06:12,626 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-07 05:06:19,480 INFO [optim.py:369] (0/4) 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 05:06:28,063 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=179806.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 05:06:31,261 WARNING [train.py:1067] (0/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-07 05:06:34,182 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179814.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:06:36,269 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5878, 1.4023, 2.7956, 1.3317, 2.1995, 3.0056, 3.2122, 2.4946], + device='cuda:0'), covar=tensor([0.1346, 0.1839, 0.0424, 0.2316, 0.0958, 0.0344, 0.0661, 0.0691], + device='cuda:0'), in_proj_covar=tensor([0.0297, 0.0322, 0.0289, 0.0316, 0.0312, 0.0269, 0.0426, 0.0307], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 05:06:41,622 INFO [train.py:901] (0/4) Epoch 23, batch 2000, loss[loss=0.1837, simple_loss=0.2766, pruned_loss=0.04539, over 7803.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2868, pruned_loss=0.06033, over 1618119.57 frames. ], batch size: 20, lr: 3.30e-03, grad_scale: 8.0 +2023-02-07 05:06:50,605 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179838.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:07:02,093 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4065, 1.2899, 2.3833, 1.2771, 2.2442, 2.5436, 2.7116, 2.1179], + device='cuda:0'), covar=tensor([0.1111, 0.1422, 0.0423, 0.2064, 0.0729, 0.0361, 0.0516, 0.0671], + device='cuda:0'), in_proj_covar=tensor([0.0296, 0.0320, 0.0287, 0.0315, 0.0311, 0.0268, 0.0424, 0.0306], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 05:07:13,192 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2602, 1.3661, 1.6699, 1.3569, 0.7522, 1.4528, 1.2905, 1.1252], + device='cuda:0'), covar=tensor([0.0558, 0.1238, 0.1598, 0.1349, 0.0545, 0.1405, 0.0647, 0.0683], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0152, 0.0189, 0.0159, 0.0099, 0.0162, 0.0111, 0.0143], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-07 05:07:16,425 INFO [train.py:901] (0/4) Epoch 23, batch 2050, loss[loss=0.1868, simple_loss=0.276, pruned_loss=0.04883, over 8465.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2865, pruned_loss=0.06016, over 1619278.25 frames. ], batch size: 25, lr: 3.30e-03, grad_scale: 8.0 +2023-02-07 05:07:30,043 INFO [optim.py:369] (0/4) 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 05:07:49,595 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179921.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 05:07:51,629 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179924.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 05:07:52,172 INFO [train.py:901] (0/4) Epoch 23, batch 2100, loss[loss=0.2126, simple_loss=0.3095, pruned_loss=0.05779, over 8462.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2869, pruned_loss=0.06052, over 1616018.63 frames. ], batch size: 25, lr: 3.30e-03, grad_scale: 8.0 +2023-02-07 05:08:04,839 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179942.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:08:14,748 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 +2023-02-07 05:08:27,545 INFO [train.py:901] (0/4) Epoch 23, batch 2150, loss[loss=0.1803, simple_loss=0.2544, pruned_loss=0.0531, over 7258.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2869, pruned_loss=0.06118, over 1612742.96 frames. ], batch size: 16, lr: 3.30e-03, grad_scale: 8.0 +2023-02-07 05:08:34,129 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6659, 2.3934, 3.7889, 1.4915, 2.8837, 2.1729, 1.9126, 2.5546], + device='cuda:0'), covar=tensor([0.2151, 0.2572, 0.1017, 0.5000, 0.1936, 0.3527, 0.2499, 0.2925], + device='cuda:0'), in_proj_covar=tensor([0.0525, 0.0608, 0.0555, 0.0645, 0.0649, 0.0592, 0.0537, 0.0630], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 05:08:34,285 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.94 vs. limit=5.0 +2023-02-07 05:08:41,580 INFO [optim.py:369] (0/4) 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 05:08:44,582 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=179998.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:08:45,954 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-180000.pt +2023-02-07 05:09:05,717 INFO [train.py:901] (0/4) Epoch 23, batch 2200, loss[loss=0.2053, simple_loss=0.2864, pruned_loss=0.06203, over 7651.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2865, pruned_loss=0.06123, over 1611965.12 frames. ], batch size: 19, lr: 3.30e-03, grad_scale: 8.0 +2023-02-07 05:09:18,801 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180044.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:09:40,653 INFO [train.py:901] (0/4) Epoch 23, batch 2250, loss[loss=0.1897, simple_loss=0.2744, pruned_loss=0.05249, over 8454.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2864, pruned_loss=0.06109, over 1616853.66 frames. ], batch size: 25, lr: 3.30e-03, grad_scale: 8.0 +2023-02-07 05:09:43,939 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-07 05:09:53,771 INFO [optim.py:369] (0/4) 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] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180094.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:10:07,878 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180113.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:10:12,137 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180119.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:10:16,772 INFO [train.py:901] (0/4) Epoch 23, batch 2300, loss[loss=0.247, simple_loss=0.3309, pruned_loss=0.08152, over 8593.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2881, pruned_loss=0.06188, over 1620270.31 frames. ], batch size: 31, lr: 3.30e-03, grad_scale: 8.0 +2023-02-07 05:10:40,106 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=180158.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:10:44,655 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 +2023-02-07 05:10:52,691 INFO [train.py:901] (0/4) Epoch 23, batch 2350, loss[loss=0.1702, simple_loss=0.2538, pruned_loss=0.04328, over 7805.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2873, pruned_loss=0.06152, over 1617153.54 frames. ], batch size: 19, lr: 3.30e-03, grad_scale: 8.0 +2023-02-07 05:10:54,356 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180177.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 05:11:05,884 INFO [optim.py:369] (0/4) 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 05:11:11,596 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180202.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 05:11:23,080 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3734, 1.6631, 4.4213, 2.0080, 2.6763, 4.9884, 5.0869, 4.2990], + device='cuda:0'), covar=tensor([0.1112, 0.1859, 0.0219, 0.1922, 0.1018, 0.0164, 0.0433, 0.0555], + device='cuda:0'), in_proj_covar=tensor([0.0294, 0.0320, 0.0286, 0.0314, 0.0310, 0.0267, 0.0422, 0.0304], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 05:11:25,800 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180223.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:11:27,091 INFO [train.py:901] (0/4) Epoch 23, batch 2400, loss[loss=0.1966, simple_loss=0.2813, pruned_loss=0.0559, over 8465.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2871, pruned_loss=0.06146, over 1617748.17 frames. ], batch size: 27, lr: 3.30e-03, grad_scale: 8.0 +2023-02-07 05:11:33,034 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-02-07 05:11:59,477 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=180268.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 05:12:02,905 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180273.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:12:04,122 INFO [train.py:901] (0/4) Epoch 23, batch 2450, loss[loss=0.2475, simple_loss=0.3232, pruned_loss=0.08594, over 7813.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2865, pruned_loss=0.061, over 1614074.04 frames. ], batch size: 20, lr: 3.30e-03, grad_scale: 8.0 +2023-02-07 05:12:12,628 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=180286.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:12:18,001 INFO [optim.py:369] (0/4) 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 05:12:39,628 INFO [train.py:901] (0/4) Epoch 23, batch 2500, loss[loss=0.2264, simple_loss=0.3103, pruned_loss=0.07119, over 8185.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2862, pruned_loss=0.06082, over 1608594.51 frames. ], batch size: 23, lr: 3.30e-03, grad_scale: 8.0 +2023-02-07 05:13:00,473 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180354.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:13:12,925 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180369.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:13:16,779 INFO [train.py:901] (0/4) Epoch 23, batch 2550, loss[loss=0.2009, simple_loss=0.275, pruned_loss=0.06343, over 7929.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2858, pruned_loss=0.06129, over 1604031.29 frames. ], batch size: 20, lr: 3.30e-03, grad_scale: 8.0 +2023-02-07 05:13:22,401 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180383.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 05:13:25,748 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=180388.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:13:29,883 INFO [optim.py:369] (0/4) 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 05:13:30,124 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180394.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:13:35,739 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180401.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:13:51,963 INFO [train.py:901] (0/4) Epoch 23, batch 2600, loss[loss=0.2784, simple_loss=0.3409, pruned_loss=0.1079, over 7558.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2866, pruned_loss=0.06201, over 1607186.25 frames. ], batch size: 72, lr: 3.30e-03, grad_scale: 8.0 +2023-02-07 05:14:28,394 INFO [train.py:901] (0/4) Epoch 23, batch 2650, loss[loss=0.1848, simple_loss=0.2847, pruned_loss=0.04244, over 8471.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2863, pruned_loss=0.06109, over 1610564.16 frames. ], batch size: 25, lr: 3.30e-03, grad_scale: 8.0 +2023-02-07 05:14:42,175 INFO [optim.py:369] (0/4) 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 05:14:48,431 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180503.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:15:04,093 INFO [train.py:901] (0/4) Epoch 23, batch 2700, loss[loss=0.1972, simple_loss=0.2875, pruned_loss=0.05345, over 8496.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2865, pruned_loss=0.0612, over 1607287.96 frames. ], batch size: 26, lr: 3.30e-03, grad_scale: 8.0 +2023-02-07 05:15:07,066 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180529.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:15:19,170 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.9650, 2.4042, 3.7588, 1.9217, 1.9335, 3.7225, 0.6583, 2.1306], + device='cuda:0'), covar=tensor([0.1402, 0.1343, 0.0221, 0.1676, 0.2562, 0.0230, 0.2349, 0.1374], + device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0200, 0.0130, 0.0222, 0.0270, 0.0137, 0.0170, 0.0195], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 05:15:24,158 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180554.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:15:33,266 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=180567.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:15:39,688 INFO [train.py:901] (0/4) Epoch 23, batch 2750, loss[loss=0.2021, simple_loss=0.3003, pruned_loss=0.05191, over 8463.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2868, pruned_loss=0.06115, over 1613226.10 frames. ], batch size: 25, lr: 3.30e-03, grad_scale: 8.0 +2023-02-07 05:15:48,802 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180588.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:15:53,500 INFO [optim.py:369] (0/4) 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 05:16:15,671 INFO [train.py:901] (0/4) Epoch 23, batch 2800, loss[loss=0.1718, simple_loss=0.27, pruned_loss=0.03685, over 8466.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2875, pruned_loss=0.06163, over 1613960.24 frames. ], batch size: 25, lr: 3.30e-03, grad_scale: 8.0 +2023-02-07 05:16:26,273 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180639.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 05:16:38,597 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180657.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:16:43,309 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180664.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 05:16:48,882 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5708, 1.8726, 1.9848, 1.2980, 2.2323, 1.3853, 0.6832, 1.8051], + device='cuda:0'), covar=tensor([0.0694, 0.0397, 0.0304, 0.0681, 0.0414, 0.0996, 0.0941, 0.0386], + device='cuda:0'), in_proj_covar=tensor([0.0457, 0.0395, 0.0348, 0.0446, 0.0380, 0.0535, 0.0393, 0.0422], + device='cuda:0'), out_proj_covar=tensor([1.2204e-04, 1.0332e-04, 9.1312e-05, 1.1730e-04, 9.9986e-05, 1.5054e-04, + 1.0598e-04, 1.1180e-04], device='cuda:0') +2023-02-07 05:16:50,739 INFO [train.py:901] (0/4) Epoch 23, batch 2850, loss[loss=0.1671, simple_loss=0.2522, pruned_loss=0.04098, over 7560.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2866, pruned_loss=0.06144, over 1612399.82 frames. ], batch size: 18, lr: 3.30e-03, grad_scale: 8.0 +2023-02-07 05:16:55,722 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180682.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:16:55,747 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180682.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:17:04,498 INFO [optim.py:369] (0/4) 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 05:17:07,424 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=180698.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:17:27,392 INFO [train.py:901] (0/4) Epoch 23, batch 2900, loss[loss=0.1953, simple_loss=0.2687, pruned_loss=0.06093, over 7976.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2871, pruned_loss=0.06163, over 1610652.04 frames. ], batch size: 21, lr: 3.30e-03, grad_scale: 8.0 +2023-02-07 05:17:45,123 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180750.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:17:52,131 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180759.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:17:59,484 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 23, batch 2950, loss[loss=0.182, simple_loss=0.2698, pruned_loss=0.04713, over 8081.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2869, pruned_loss=0.06155, over 1608616.47 frames. ], batch size: 21, lr: 3.30e-03, grad_scale: 8.0 +2023-02-07 05:18:09,318 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180784.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:18:16,022 INFO [optim.py:369] (0/4) 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 05:18:26,739 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1549, 1.6486, 1.7968, 1.5266, 1.0157, 1.5542, 1.9817, 1.6354], + device='cuda:0'), covar=tensor([0.0481, 0.1112, 0.1578, 0.1325, 0.0576, 0.1357, 0.0595, 0.0618], + device='cuda:0'), in_proj_covar=tensor([0.0096, 0.0152, 0.0188, 0.0159, 0.0099, 0.0161, 0.0111, 0.0142], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-07 05:18:30,321 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180813.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:18:38,129 INFO [train.py:901] (0/4) Epoch 23, batch 3000, loss[loss=0.2009, simple_loss=0.2904, pruned_loss=0.05568, over 8336.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2871, pruned_loss=0.06114, over 1611415.09 frames. ], batch size: 25, lr: 3.29e-03, grad_scale: 8.0 +2023-02-07 05:18:38,130 INFO [train.py:926] (0/4) Computing validation loss +2023-02-07 05:18:45,773 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2870, 1.0900, 1.2608, 1.0262, 0.6815, 1.0958, 1.1932, 1.0161], + device='cuda:0'), covar=tensor([0.0459, 0.1044, 0.1303, 0.1156, 0.0514, 0.1187, 0.0527, 0.0524], + device='cuda:0'), in_proj_covar=tensor([0.0096, 0.0152, 0.0189, 0.0159, 0.0099, 0.0161, 0.0111, 0.0142], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-07 05:18:50,537 INFO [train.py:935] (0/4) 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] (0/4) Maximum memory allocated so far is 6722MB +2023-02-07 05:19:03,682 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180843.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:19:26,992 INFO [train.py:901] (0/4) Epoch 23, batch 3050, loss[loss=0.2138, simple_loss=0.2965, pruned_loss=0.06555, over 8146.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2869, pruned_loss=0.06149, over 1610169.02 frames. ], batch size: 48, lr: 3.29e-03, grad_scale: 8.0 +2023-02-07 05:19:40,669 INFO [optim.py:369] (0/4) 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 05:20:02,333 INFO [train.py:901] (0/4) Epoch 23, batch 3100, loss[loss=0.1902, simple_loss=0.2599, pruned_loss=0.06022, over 7691.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2867, pruned_loss=0.06111, over 1611342.01 frames. ], batch size: 18, lr: 3.29e-03, grad_scale: 8.0 +2023-02-07 05:20:07,174 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=180932.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:20:11,431 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180938.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:20:22,569 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5583, 1.7980, 1.9228, 1.4092, 1.9857, 1.4243, 0.4459, 1.7924], + device='cuda:0'), covar=tensor([0.0526, 0.0342, 0.0294, 0.0488, 0.0404, 0.0947, 0.0896, 0.0251], + device='cuda:0'), in_proj_covar=tensor([0.0457, 0.0395, 0.0347, 0.0448, 0.0380, 0.0536, 0.0393, 0.0422], + device='cuda:0'), out_proj_covar=tensor([1.2212e-04, 1.0333e-04, 9.1140e-05, 1.1778e-04, 1.0004e-04, 1.5101e-04, + 1.0585e-04, 1.1162e-04], device='cuda:0') +2023-02-07 05:20:29,335 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 23, batch 3150, loss[loss=0.2127, simple_loss=0.2921, pruned_loss=0.06662, over 7652.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2873, pruned_loss=0.06155, over 1613576.27 frames. ], batch size: 19, lr: 3.29e-03, grad_scale: 8.0 +2023-02-07 05:20:51,967 INFO [optim.py:369] (0/4) 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 05:21:14,471 INFO [train.py:901] (0/4) Epoch 23, batch 3200, loss[loss=0.1883, simple_loss=0.2715, pruned_loss=0.0525, over 8105.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2879, pruned_loss=0.06185, over 1613284.66 frames. ], batch size: 23, lr: 3.29e-03, grad_scale: 8.0 +2023-02-07 05:21:29,909 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=181047.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:21:45,848 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=181069.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:21:49,882 INFO [train.py:901] (0/4) Epoch 23, batch 3250, loss[loss=0.1934, simple_loss=0.2794, pruned_loss=0.05372, over 7802.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2884, pruned_loss=0.06245, over 1608804.35 frames. ], batch size: 20, lr: 3.29e-03, grad_scale: 8.0 +2023-02-07 05:22:03,761 INFO [optim.py:369] (0/4) 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 05:22:03,868 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=181094.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:22:03,996 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=181094.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:22:25,755 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4246, 1.4022, 1.8097, 1.3177, 1.0889, 1.7893, 0.1897, 1.1799], + device='cuda:0'), covar=tensor([0.1970, 0.1368, 0.0435, 0.0969, 0.2892, 0.0473, 0.2263, 0.1280], + device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0199, 0.0130, 0.0222, 0.0272, 0.0138, 0.0171, 0.0194], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 05:22:26,226 INFO [train.py:901] (0/4) Epoch 23, batch 3300, loss[loss=0.2136, simple_loss=0.3006, pruned_loss=0.06328, over 8449.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2871, pruned_loss=0.06177, over 1604176.48 frames. ], batch size: 27, lr: 3.29e-03, grad_scale: 8.0 +2023-02-07 05:23:01,363 INFO [train.py:901] (0/4) Epoch 23, batch 3350, loss[loss=0.2369, simple_loss=0.3186, pruned_loss=0.07755, over 8425.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2866, pruned_loss=0.06115, over 1605177.59 frames. ], batch size: 27, lr: 3.29e-03, grad_scale: 8.0 +2023-02-07 05:23:10,452 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 05:23:26,406 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=181209.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:23:38,218 INFO [train.py:901] (0/4) Epoch 23, batch 3400, loss[loss=0.1878, simple_loss=0.2771, pruned_loss=0.0493, over 8237.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2873, pruned_loss=0.06166, over 1604316.85 frames. ], batch size: 22, lr: 3.29e-03, grad_scale: 8.0 +2023-02-07 05:24:13,225 INFO [train.py:901] (0/4) Epoch 23, batch 3450, loss[loss=0.2091, simple_loss=0.2766, pruned_loss=0.07074, over 7407.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2876, pruned_loss=0.06189, over 1607451.74 frames. ], batch size: 17, lr: 3.29e-03, grad_scale: 8.0 +2023-02-07 05:24:27,415 INFO [optim.py:369] (0/4) 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 05:24:32,992 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=181302.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:24:33,727 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=181303.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:24:42,773 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=181315.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:24:47,013 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6753, 2.7547, 1.9788, 2.4602, 2.3534, 1.7728, 2.2457, 2.3726], + device='cuda:0'), covar=tensor([0.1453, 0.0386, 0.1170, 0.0604, 0.0732, 0.1458, 0.1002, 0.0956], + device='cuda:0'), in_proj_covar=tensor([0.0356, 0.0233, 0.0336, 0.0309, 0.0300, 0.0338, 0.0344, 0.0316], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 05:24:49,555 INFO [train.py:901] (0/4) Epoch 23, batch 3500, loss[loss=0.192, simple_loss=0.2805, pruned_loss=0.05182, over 8100.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2866, pruned_loss=0.06108, over 1610786.90 frames. ], batch size: 23, lr: 3.29e-03, grad_scale: 8.0 +2023-02-07 05:24:52,789 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=181328.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:25:07,647 WARNING [train.py:1067] (0/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-07 05:25:13,730 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.60 vs. limit=5.0 +2023-02-07 05:25:17,763 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 +2023-02-07 05:25:25,788 INFO [train.py:901] (0/4) Epoch 23, batch 3550, loss[loss=0.1934, simple_loss=0.2862, pruned_loss=0.05037, over 8090.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2859, pruned_loss=0.06062, over 1611393.74 frames. ], batch size: 21, lr: 3.29e-03, grad_scale: 8.0 +2023-02-07 05:25:39,019 INFO [optim.py:369] (0/4) 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 05:26:01,193 INFO [train.py:901] (0/4) Epoch 23, batch 3600, loss[loss=0.2027, simple_loss=0.2752, pruned_loss=0.06513, over 7653.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2872, pruned_loss=0.0612, over 1617167.49 frames. ], batch size: 19, lr: 3.29e-03, grad_scale: 8.0 +2023-02-07 05:26:09,774 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4632, 1.8615, 3.2022, 1.3205, 2.4754, 1.9914, 1.5295, 2.4247], + device='cuda:0'), covar=tensor([0.2222, 0.2875, 0.0813, 0.5026, 0.1844, 0.3347, 0.2711, 0.2137], + device='cuda:0'), in_proj_covar=tensor([0.0524, 0.0604, 0.0549, 0.0643, 0.0644, 0.0591, 0.0536, 0.0627], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 05:26:30,524 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=181465.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:26:37,794 INFO [train.py:901] (0/4) Epoch 23, batch 3650, loss[loss=0.1752, simple_loss=0.2626, pruned_loss=0.04386, over 7827.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2865, pruned_loss=0.06062, over 1618686.60 frames. ], batch size: 20, lr: 3.29e-03, grad_scale: 8.0 +2023-02-07 05:26:44,025 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1398, 1.5828, 3.4530, 1.6177, 2.3713, 3.8827, 3.9566, 3.2898], + device='cuda:0'), covar=tensor([0.1082, 0.1868, 0.0344, 0.2075, 0.1158, 0.0220, 0.0475, 0.0570], + device='cuda:0'), in_proj_covar=tensor([0.0295, 0.0319, 0.0286, 0.0314, 0.0309, 0.0267, 0.0422, 0.0303], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 05:26:48,286 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=181490.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:26:50,927 INFO [optim.py:369] (0/4) 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 05:26:52,533 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.5712, 4.5937, 4.1383, 2.1422, 4.0577, 4.1819, 4.0789, 4.0194], + device='cuda:0'), covar=tensor([0.0715, 0.0526, 0.1133, 0.4729, 0.0861, 0.1080, 0.1299, 0.0755], + device='cuda:0'), in_proj_covar=tensor([0.0527, 0.0439, 0.0431, 0.0540, 0.0429, 0.0444, 0.0426, 0.0389], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 05:27:11,129 WARNING [train.py:1067] (0/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-07 05:27:12,456 INFO [train.py:901] (0/4) Epoch 23, batch 3700, loss[loss=0.2423, simple_loss=0.3186, pruned_loss=0.08301, over 8251.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2872, pruned_loss=0.06113, over 1615926.05 frames. ], batch size: 24, lr: 3.29e-03, grad_scale: 8.0 +2023-02-07 05:27:30,273 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=181548.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:27:37,393 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=181558.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:27:49,547 INFO [train.py:901] (0/4) Epoch 23, batch 3750, loss[loss=0.2056, simple_loss=0.2827, pruned_loss=0.06425, over 7946.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.287, pruned_loss=0.06107, over 1617227.62 frames. ], batch size: 20, lr: 3.29e-03, grad_scale: 16.0 +2023-02-07 05:27:55,369 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=181583.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:28:02,819 INFO [optim.py:369] (0/4) 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 05:28:09,865 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.3956, 1.5900, 4.5966, 1.6173, 4.0508, 3.7612, 4.1281, 4.0178], + device='cuda:0'), covar=tensor([0.0608, 0.4421, 0.0467, 0.4200, 0.1131, 0.0931, 0.0551, 0.0675], + device='cuda:0'), in_proj_covar=tensor([0.0644, 0.0651, 0.0707, 0.0641, 0.0722, 0.0617, 0.0617, 0.0689], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 05:28:24,873 INFO [train.py:901] (0/4) Epoch 23, batch 3800, loss[loss=0.1867, simple_loss=0.2752, pruned_loss=0.04914, over 8099.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2861, pruned_loss=0.06087, over 1610322.85 frames. ], batch size: 21, lr: 3.29e-03, grad_scale: 16.0 +2023-02-07 05:28:44,966 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9387, 1.4404, 3.5427, 1.5598, 2.3907, 3.9351, 4.0036, 3.4114], + device='cuda:0'), covar=tensor([0.1145, 0.1852, 0.0273, 0.2035, 0.1007, 0.0193, 0.0376, 0.0486], + device='cuda:0'), in_proj_covar=tensor([0.0296, 0.0320, 0.0285, 0.0315, 0.0309, 0.0267, 0.0421, 0.0303], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 05:28:49,194 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=181659.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:29:00,796 INFO [train.py:901] (0/4) Epoch 23, batch 3850, loss[loss=0.2366, simple_loss=0.3017, pruned_loss=0.08574, over 8652.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2872, pruned_loss=0.06172, over 1613502.44 frames. ], batch size: 34, lr: 3.29e-03, grad_scale: 16.0 +2023-02-07 05:29:14,852 INFO [optim.py:369] (0/4) 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 05:29:15,140 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6865, 2.3533, 4.0986, 1.5688, 2.9027, 2.2136, 1.8876, 2.7320], + device='cuda:0'), covar=tensor([0.2047, 0.2644, 0.0760, 0.4750, 0.1895, 0.3354, 0.2449, 0.2564], + device='cuda:0'), in_proj_covar=tensor([0.0527, 0.0609, 0.0554, 0.0649, 0.0650, 0.0599, 0.0542, 0.0633], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 05:29:17,199 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4711, 1.4210, 1.8489, 1.1019, 1.0418, 1.8074, 0.1171, 1.1131], + device='cuda:0'), covar=tensor([0.1582, 0.1319, 0.0381, 0.1212, 0.2714, 0.0433, 0.2009, 0.1224], + device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0199, 0.0128, 0.0219, 0.0269, 0.0136, 0.0170, 0.0192], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 05:29:22,392 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-07 05:29:36,640 INFO [train.py:901] (0/4) Epoch 23, batch 3900, loss[loss=0.195, simple_loss=0.2846, pruned_loss=0.05275, over 8479.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2872, pruned_loss=0.06129, over 1613695.47 frames. ], batch size: 29, lr: 3.29e-03, grad_scale: 16.0 +2023-02-07 05:30:10,464 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 23, batch 3950, loss[loss=0.2063, simple_loss=0.2992, pruned_loss=0.05668, over 8556.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2879, pruned_loss=0.0618, over 1614178.52 frames. ], batch size: 31, lr: 3.29e-03, grad_scale: 16.0 +2023-02-07 05:30:26,260 INFO [optim.py:369] (0/4) 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 05:30:47,713 INFO [train.py:901] (0/4) Epoch 23, batch 4000, loss[loss=0.1934, simple_loss=0.2805, pruned_loss=0.05313, over 8253.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2881, pruned_loss=0.06153, over 1616522.28 frames. ], batch size: 24, lr: 3.29e-03, grad_scale: 16.0 +2023-02-07 05:31:22,578 INFO [train.py:901] (0/4) Epoch 23, batch 4050, loss[loss=0.2454, simple_loss=0.3254, pruned_loss=0.08266, over 8442.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2877, pruned_loss=0.06185, over 1617396.26 frames. ], batch size: 49, lr: 3.29e-03, grad_scale: 16.0 +2023-02-07 05:31:34,361 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=181892.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:31:34,878 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 +2023-02-07 05:31:35,719 INFO [optim.py:369] (0/4) 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 05:31:59,840 INFO [train.py:901] (0/4) Epoch 23, batch 4100, loss[loss=0.2185, simple_loss=0.3019, pruned_loss=0.06753, over 8374.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2881, pruned_loss=0.06154, over 1619005.31 frames. ], batch size: 24, lr: 3.29e-03, grad_scale: 16.0 +2023-02-07 05:32:34,923 INFO [train.py:901] (0/4) Epoch 23, batch 4150, loss[loss=0.2198, simple_loss=0.3108, pruned_loss=0.06445, over 8342.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2905, pruned_loss=0.06294, over 1618994.40 frames. ], batch size: 26, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:32:48,434 INFO [optim.py:369] (0/4) 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 05:32:49,356 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=181996.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:32:52,060 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-182000.pt +2023-02-07 05:32:58,727 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=182007.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:33:11,748 INFO [train.py:901] (0/4) Epoch 23, batch 4200, loss[loss=0.16, simple_loss=0.2497, pruned_loss=0.03517, over 7975.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2891, pruned_loss=0.06237, over 1614232.43 frames. ], batch size: 21, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:33:16,172 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=182030.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:33:25,741 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-07 05:33:33,435 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=182055.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:33:43,683 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9960, 1.3500, 1.5926, 1.2917, 0.8826, 1.4364, 1.7097, 1.4635], + device='cuda:0'), covar=tensor([0.0528, 0.1361, 0.1754, 0.1511, 0.0616, 0.1568, 0.0715, 0.0697], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0152, 0.0190, 0.0160, 0.0100, 0.0163, 0.0112, 0.0143], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-07 05:33:47,604 INFO [train.py:901] (0/4) Epoch 23, batch 4250, loss[loss=0.2357, simple_loss=0.3158, pruned_loss=0.07778, over 8613.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2881, pruned_loss=0.06166, over 1612952.96 frames. ], batch size: 34, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:33:49,061 WARNING [train.py:1067] (0/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-07 05:34:01,341 INFO [optim.py:369] (0/4) 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 05:34:22,743 INFO [train.py:901] (0/4) Epoch 23, batch 4300, loss[loss=0.1978, simple_loss=0.2915, pruned_loss=0.05208, over 8240.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2874, pruned_loss=0.06152, over 1609950.42 frames. ], batch size: 24, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:34:58,238 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7445, 1.8006, 2.4041, 1.5815, 1.4286, 2.3492, 0.3275, 1.4358], + device='cuda:0'), covar=tensor([0.1538, 0.1049, 0.0268, 0.1018, 0.2218, 0.0320, 0.1878, 0.1042], + device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0200, 0.0129, 0.0221, 0.0270, 0.0137, 0.0170, 0.0193], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 05:34:58,707 INFO [train.py:901] (0/4) Epoch 23, batch 4350, loss[loss=0.193, simple_loss=0.2702, pruned_loss=0.05792, over 7563.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2869, pruned_loss=0.06139, over 1609799.38 frames. ], batch size: 18, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:35:13,491 INFO [optim.py:369] (0/4) 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 05:35:21,947 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-07 05:35:34,673 INFO [train.py:901] (0/4) Epoch 23, batch 4400, loss[loss=0.2061, simple_loss=0.2903, pruned_loss=0.06098, over 8505.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.286, pruned_loss=0.06076, over 1611188.50 frames. ], batch size: 26, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:36:01,785 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9951, 2.2535, 1.8745, 2.9294, 1.3607, 1.6852, 2.1669, 2.3685], + device='cuda:0'), covar=tensor([0.0718, 0.0789, 0.0824, 0.0389, 0.1126, 0.1217, 0.0814, 0.0767], + device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0198, 0.0246, 0.0215, 0.0208, 0.0249, 0.0251, 0.0209], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 05:36:03,219 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=182263.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:36:05,078 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-07 05:36:11,440 INFO [train.py:901] (0/4) Epoch 23, batch 4450, loss[loss=0.2236, simple_loss=0.2991, pruned_loss=0.07406, over 8450.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2857, pruned_loss=0.06045, over 1611924.55 frames. ], batch size: 27, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:36:20,465 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=182288.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:36:26,040 INFO [optim.py:369] (0/4) 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 05:36:28,971 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=182299.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:36:47,069 INFO [train.py:901] (0/4) Epoch 23, batch 4500, loss[loss=0.2433, simple_loss=0.3237, pruned_loss=0.08151, over 8529.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.286, pruned_loss=0.06079, over 1612948.37 frames. ], batch size: 49, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:36:56,821 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-07 05:36:57,583 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=182340.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:37:23,620 INFO [train.py:901] (0/4) Epoch 23, batch 4550, loss[loss=0.2604, simple_loss=0.3203, pruned_loss=0.1003, over 7705.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2855, pruned_loss=0.0604, over 1609507.32 frames. ], batch size: 18, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:37:37,490 INFO [optim.py:369] (0/4) 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 05:37:39,975 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 +2023-02-07 05:37:59,264 INFO [train.py:901] (0/4) Epoch 23, batch 4600, loss[loss=0.189, simple_loss=0.2667, pruned_loss=0.05566, over 8087.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2857, pruned_loss=0.06028, over 1611532.75 frames. ], batch size: 21, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:38:20,338 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=182455.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:38:34,862 INFO [train.py:901] (0/4) Epoch 23, batch 4650, loss[loss=0.2081, simple_loss=0.2858, pruned_loss=0.06521, over 8080.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2863, pruned_loss=0.06076, over 1614244.90 frames. ], batch size: 21, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:38:50,632 INFO [optim.py:369] (0/4) 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 05:39:12,445 INFO [train.py:901] (0/4) Epoch 23, batch 4700, loss[loss=0.1803, simple_loss=0.2799, pruned_loss=0.04035, over 8255.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2856, pruned_loss=0.05997, over 1617069.91 frames. ], batch size: 24, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:39:17,385 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=182532.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:39:47,033 INFO [train.py:901] (0/4) Epoch 23, batch 4750, loss[loss=0.1968, simple_loss=0.2942, pruned_loss=0.04972, over 8337.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2858, pruned_loss=0.05974, over 1619389.13 frames. ], batch size: 26, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:40:01,607 INFO [optim.py:369] (0/4) 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 05:40:06,065 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-07 05:40:08,822 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-07 05:40:24,084 INFO [train.py:901] (0/4) Epoch 23, batch 4800, loss[loss=0.2171, simple_loss=0.297, pruned_loss=0.06862, over 8047.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2855, pruned_loss=0.05953, over 1616051.11 frames. ], batch size: 22, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:40:36,435 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=182643.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:40:59,186 INFO [train.py:901] (0/4) Epoch 23, batch 4850, loss[loss=0.2231, simple_loss=0.293, pruned_loss=0.0766, over 8334.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2867, pruned_loss=0.0606, over 1614903.51 frames. ], batch size: 26, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:41:00,607 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-07 05:41:13,249 INFO [optim.py:369] (0/4) 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 05:41:25,682 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=182711.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:41:36,257 INFO [train.py:901] (0/4) Epoch 23, batch 4900, loss[loss=0.1811, simple_loss=0.2666, pruned_loss=0.04782, over 8187.00 frames. ], tot_loss[loss=0.203, simple_loss=0.286, pruned_loss=0.05995, over 1617571.27 frames. ], batch size: 23, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:41:44,985 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=182736.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:42:00,347 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=182758.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:42:12,960 INFO [train.py:901] (0/4) Epoch 23, batch 4950, loss[loss=0.1946, simple_loss=0.2727, pruned_loss=0.05831, over 7519.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2864, pruned_loss=0.06044, over 1616546.86 frames. ], batch size: 18, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:42:27,040 INFO [optim.py:369] (0/4) 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 05:42:48,225 INFO [train.py:901] (0/4) Epoch 23, batch 5000, loss[loss=0.2617, simple_loss=0.3229, pruned_loss=0.1003, over 6722.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.286, pruned_loss=0.0607, over 1613007.74 frames. ], batch size: 71, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:43:25,214 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=182874.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:43:25,725 INFO [train.py:901] (0/4) Epoch 23, batch 5050, loss[loss=0.1815, simple_loss=0.2621, pruned_loss=0.05044, over 7819.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2864, pruned_loss=0.06117, over 1610756.14 frames. ], batch size: 20, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:43:26,529 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=182876.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:43:29,514 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=182880.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:43:40,730 INFO [optim.py:369] (0/4) 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 05:43:46,341 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-07 05:44:01,709 INFO [train.py:901] (0/4) Epoch 23, batch 5100, loss[loss=0.2299, simple_loss=0.3075, pruned_loss=0.07618, over 8502.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2865, pruned_loss=0.06116, over 1606214.23 frames. ], batch size: 28, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:44:03,885 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9673, 1.4836, 1.5793, 1.3771, 0.9551, 1.3858, 1.6746, 1.6200], + device='cuda:0'), covar=tensor([0.0553, 0.1284, 0.1757, 0.1496, 0.0619, 0.1538, 0.0736, 0.0646], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0153, 0.0190, 0.0160, 0.0101, 0.0163, 0.0112, 0.0144], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-07 05:44:38,859 INFO [train.py:901] (0/4) Epoch 23, batch 5150, loss[loss=0.1878, simple_loss=0.2813, pruned_loss=0.04716, over 8361.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2876, pruned_loss=0.06152, over 1611675.03 frames. ], batch size: 24, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:44:50,234 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=182991.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:44:53,603 INFO [optim.py:369] (0/4) 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 05:45:07,238 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=183014.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:45:09,899 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.6099, 1.5830, 1.8859, 1.7268, 1.0608, 1.6850, 2.1600, 2.0262], + device='cuda:0'), covar=tensor([0.0505, 0.1247, 0.1554, 0.1323, 0.0613, 0.1431, 0.0630, 0.0570], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0153, 0.0189, 0.0160, 0.0101, 0.0162, 0.0112, 0.0143], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0007], + device='cuda:0') +2023-02-07 05:45:14,591 INFO [train.py:901] (0/4) Epoch 23, batch 5200, loss[loss=0.1947, simple_loss=0.2668, pruned_loss=0.06132, over 7426.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2869, pruned_loss=0.0607, over 1612321.66 frames. ], batch size: 17, lr: 3.28e-03, grad_scale: 8.0 +2023-02-07 05:45:24,441 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=183039.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:45:46,581 WARNING [train.py:1067] (0/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-07 05:45:50,599 INFO [train.py:901] (0/4) Epoch 23, batch 5250, loss[loss=0.2251, simple_loss=0.3017, pruned_loss=0.07421, over 8462.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2882, pruned_loss=0.06133, over 1618095.56 frames. ], batch size: 25, lr: 3.27e-03, grad_scale: 8.0 +2023-02-07 05:45:55,149 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 +2023-02-07 05:46:05,168 INFO [optim.py:369] (0/4) 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 05:46:27,067 INFO [train.py:901] (0/4) Epoch 23, batch 5300, loss[loss=0.2297, simple_loss=0.3057, pruned_loss=0.07684, over 7517.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2872, pruned_loss=0.06074, over 1617417.66 frames. ], batch size: 71, lr: 3.27e-03, grad_scale: 8.0 +2023-02-07 05:46:32,349 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.68 vs. limit=5.0 +2023-02-07 05:46:59,017 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.59 vs. limit=5.0 +2023-02-07 05:47:02,886 INFO [train.py:901] (0/4) Epoch 23, batch 5350, loss[loss=0.2472, simple_loss=0.3202, pruned_loss=0.08711, over 8467.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.288, pruned_loss=0.06089, over 1618292.30 frames. ], batch size: 25, lr: 3.27e-03, grad_scale: 8.0 +2023-02-07 05:47:17,709 INFO [optim.py:369] (0/4) 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 05:47:34,655 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=183218.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:47:38,841 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=183224.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:47:39,486 INFO [train.py:901] (0/4) Epoch 23, batch 5400, loss[loss=0.2207, simple_loss=0.3043, pruned_loss=0.06852, over 8206.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2869, pruned_loss=0.06035, over 1615160.31 frames. ], batch size: 23, lr: 3.27e-03, grad_scale: 4.0 +2023-02-07 05:47:55,891 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=183247.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:48:13,007 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=183272.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:48:14,973 INFO [train.py:901] (0/4) Epoch 23, batch 5450, loss[loss=0.1596, simple_loss=0.2361, pruned_loss=0.04149, over 7515.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2871, pruned_loss=0.06042, over 1620094.05 frames. ], batch size: 18, lr: 3.27e-03, grad_scale: 4.0 +2023-02-07 05:48:22,225 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1193, 1.9052, 2.5002, 2.1284, 2.5319, 2.1241, 1.9348, 1.4261], + device='cuda:0'), covar=tensor([0.5492, 0.4965, 0.1974, 0.3568, 0.2173, 0.3237, 0.2063, 0.5171], + device='cuda:0'), in_proj_covar=tensor([0.0941, 0.0986, 0.0806, 0.0947, 0.0995, 0.0898, 0.0751, 0.0829], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 05:48:30,412 INFO [optim.py:369] (0/4) 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 05:48:41,150 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 23, batch 5500, loss[loss=0.2035, simple_loss=0.2903, pruned_loss=0.05837, over 8508.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2879, pruned_loss=0.06079, over 1617618.47 frames. ], batch size: 39, lr: 3.27e-03, grad_scale: 4.0 +2023-02-07 05:48:58,314 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=183333.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:49:02,343 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=183339.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:49:14,869 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 +2023-02-07 05:49:27,066 INFO [train.py:901] (0/4) Epoch 23, batch 5550, loss[loss=0.1914, simple_loss=0.2878, pruned_loss=0.04752, over 8485.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2873, pruned_loss=0.06068, over 1612594.01 frames. ], batch size: 28, lr: 3.27e-03, grad_scale: 4.0 +2023-02-07 05:49:41,518 INFO [optim.py:369] (0/4) 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 05:50:03,270 INFO [train.py:901] (0/4) Epoch 23, batch 5600, loss[loss=0.1709, simple_loss=0.2571, pruned_loss=0.04233, over 7929.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2865, pruned_loss=0.06045, over 1611480.62 frames. ], batch size: 20, lr: 3.27e-03, grad_scale: 8.0 +2023-02-07 05:50:04,075 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=183426.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:50:12,353 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.3677, 1.6272, 4.6112, 1.7475, 4.0676, 3.8272, 4.1701, 4.0701], + device='cuda:0'), covar=tensor([0.0625, 0.4221, 0.0498, 0.3728, 0.1068, 0.0881, 0.0543, 0.0628], + device='cuda:0'), in_proj_covar=tensor([0.0643, 0.0650, 0.0704, 0.0636, 0.0713, 0.0613, 0.0611, 0.0686], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 05:50:39,034 INFO [train.py:901] (0/4) Epoch 23, batch 5650, loss[loss=0.2156, simple_loss=0.2929, pruned_loss=0.06911, over 7918.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2855, pruned_loss=0.06025, over 1609924.84 frames. ], batch size: 20, lr: 3.27e-03, grad_scale: 8.0 +2023-02-07 05:50:51,413 WARNING [train.py:1067] (0/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-07 05:50:53,306 INFO [optim.py:369] (0/4) 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 05:51:14,121 INFO [train.py:901] (0/4) Epoch 23, batch 5700, loss[loss=0.2288, simple_loss=0.3082, pruned_loss=0.07466, over 8644.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2867, pruned_loss=0.06104, over 1608293.79 frames. ], batch size: 39, lr: 3.27e-03, grad_scale: 8.0 +2023-02-07 05:51:50,330 INFO [train.py:901] (0/4) Epoch 23, batch 5750, loss[loss=0.2075, simple_loss=0.2919, pruned_loss=0.06155, over 8590.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2856, pruned_loss=0.06136, over 1605401.87 frames. ], batch size: 39, lr: 3.27e-03, grad_scale: 8.0 +2023-02-07 05:51:57,138 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-07 05:52:00,834 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=183589.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:52:01,859 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.85 vs. limit=5.0 +2023-02-07 05:52:05,028 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 05:52:18,045 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=183614.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:52:22,120 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=183620.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:52:25,346 INFO [train.py:901] (0/4) Epoch 23, batch 5800, loss[loss=0.1903, simple_loss=0.2784, pruned_loss=0.05112, over 8523.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2851, pruned_loss=0.06026, over 1609914.65 frames. ], batch size: 28, lr: 3.27e-03, grad_scale: 8.0 +2023-02-07 05:52:30,168 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=183632.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:52:51,758 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.3894, 1.3097, 4.6265, 1.7407, 4.1346, 3.8245, 4.1790, 4.0713], + device='cuda:0'), covar=tensor([0.0565, 0.4801, 0.0471, 0.3963, 0.0989, 0.0961, 0.0523, 0.0660], + device='cuda:0'), in_proj_covar=tensor([0.0642, 0.0650, 0.0704, 0.0636, 0.0713, 0.0614, 0.0610, 0.0689], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 05:53:01,086 INFO [train.py:901] (0/4) Epoch 23, batch 5850, loss[loss=0.1666, simple_loss=0.2514, pruned_loss=0.04087, over 7657.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2843, pruned_loss=0.05976, over 1607175.88 frames. ], batch size: 19, lr: 3.27e-03, grad_scale: 8.0 +2023-02-07 05:53:16,202 INFO [optim.py:369] (0/4) 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 05:53:36,768 INFO [train.py:901] (0/4) Epoch 23, batch 5900, loss[loss=0.215, simple_loss=0.2962, pruned_loss=0.06691, over 8524.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2841, pruned_loss=0.05923, over 1610779.33 frames. ], batch size: 28, lr: 3.27e-03, grad_scale: 8.0 +2023-02-07 05:53:38,958 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6959, 1.4648, 2.8923, 1.4276, 2.3184, 3.0854, 3.2435, 2.6023], + device='cuda:0'), covar=tensor([0.1144, 0.1613, 0.0351, 0.1996, 0.0761, 0.0302, 0.0635, 0.0585], + device='cuda:0'), in_proj_covar=tensor([0.0294, 0.0319, 0.0285, 0.0313, 0.0311, 0.0267, 0.0422, 0.0302], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 05:53:38,995 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7397, 1.8784, 1.6336, 2.3724, 1.0332, 1.4977, 1.7318, 1.9067], + device='cuda:0'), covar=tensor([0.0746, 0.0756, 0.0942, 0.0412, 0.1002, 0.1341, 0.0693, 0.0690], + device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0197, 0.0245, 0.0214, 0.0206, 0.0246, 0.0250, 0.0206], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 05:53:59,044 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 +2023-02-07 05:54:08,420 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=183770.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:54:11,756 INFO [train.py:901] (0/4) Epoch 23, batch 5950, loss[loss=0.2166, simple_loss=0.3019, pruned_loss=0.06567, over 8021.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2853, pruned_loss=0.05971, over 1610440.61 frames. ], batch size: 22, lr: 3.27e-03, grad_scale: 8.0 +2023-02-07 05:54:23,697 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.2002, 4.1755, 3.8160, 1.9479, 3.7377, 3.7577, 3.7879, 3.5736], + device='cuda:0'), covar=tensor([0.0889, 0.0635, 0.1183, 0.4312, 0.1001, 0.1147, 0.1368, 0.0854], + device='cuda:0'), in_proj_covar=tensor([0.0524, 0.0441, 0.0426, 0.0536, 0.0427, 0.0443, 0.0427, 0.0384], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 05:54:27,017 INFO [optim.py:369] (0/4) 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 05:54:38,314 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4250, 1.2940, 2.3821, 1.2593, 2.1063, 2.5292, 2.6973, 2.1470], + device='cuda:0'), covar=tensor([0.1134, 0.1474, 0.0459, 0.2091, 0.0788, 0.0417, 0.0695, 0.0671], + device='cuda:0'), in_proj_covar=tensor([0.0293, 0.0317, 0.0283, 0.0311, 0.0308, 0.0265, 0.0420, 0.0300], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 05:54:43,514 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.5855, 4.6311, 4.1449, 2.1406, 4.1330, 4.1732, 4.2071, 4.0361], + device='cuda:0'), covar=tensor([0.0728, 0.0483, 0.1092, 0.4492, 0.0808, 0.0971, 0.1179, 0.0764], + device='cuda:0'), in_proj_covar=tensor([0.0526, 0.0443, 0.0428, 0.0539, 0.0429, 0.0445, 0.0428, 0.0385], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 05:54:47,613 INFO [train.py:901] (0/4) Epoch 23, batch 6000, loss[loss=0.192, simple_loss=0.2727, pruned_loss=0.05571, over 8023.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2858, pruned_loss=0.06005, over 1613221.31 frames. ], batch size: 22, lr: 3.27e-03, grad_scale: 8.0 +2023-02-07 05:54:47,614 INFO [train.py:926] (0/4) Computing validation loss +2023-02-07 05:55:00,695 INFO [train.py:935] (0/4) Epoch 23, validation: loss=0.1722, simple_loss=0.2724, pruned_loss=0.03597, over 944034.00 frames. +2023-02-07 05:55:00,697 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-07 05:55:19,887 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.19 vs. limit=5.0 +2023-02-07 05:55:25,805 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=183860.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:55:27,316 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.7435, 2.5034, 3.2011, 2.6500, 3.0884, 2.5992, 2.5524, 2.4062], + device='cuda:0'), covar=tensor([0.3855, 0.3960, 0.1676, 0.3059, 0.1965, 0.2611, 0.1446, 0.3727], + device='cuda:0'), in_proj_covar=tensor([0.0945, 0.0990, 0.0809, 0.0953, 0.0998, 0.0900, 0.0751, 0.0829], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 05:55:36,138 INFO [train.py:901] (0/4) Epoch 23, batch 6050, loss[loss=0.2034, simple_loss=0.2646, pruned_loss=0.0711, over 7537.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.286, pruned_loss=0.06036, over 1613122.10 frames. ], batch size: 18, lr: 3.27e-03, grad_scale: 8.0 +2023-02-07 05:55:43,231 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 05:56:11,858 INFO [train.py:901] (0/4) Epoch 23, batch 6100, loss[loss=0.2142, simple_loss=0.2954, pruned_loss=0.06654, over 8625.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2875, pruned_loss=0.0615, over 1617177.12 frames. ], batch size: 34, lr: 3.27e-03, grad_scale: 8.0 +2023-02-07 05:56:21,661 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-02-07 05:56:23,562 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.4044, 4.3821, 4.0046, 2.2802, 3.9162, 3.9683, 3.9452, 3.9051], + device='cuda:0'), covar=tensor([0.0714, 0.0517, 0.0961, 0.3920, 0.0846, 0.1046, 0.1292, 0.0748], + device='cuda:0'), in_proj_covar=tensor([0.0526, 0.0443, 0.0431, 0.0539, 0.0430, 0.0446, 0.0428, 0.0386], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 05:56:32,486 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-07 05:56:47,362 INFO [train.py:901] (0/4) Epoch 23, batch 6150, loss[loss=0.1973, simple_loss=0.2809, pruned_loss=0.05683, over 7921.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.287, pruned_loss=0.06124, over 1614633.14 frames. ], batch size: 20, lr: 3.27e-03, grad_scale: 8.0 +2023-02-07 05:56:48,168 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=183976.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:56:51,714 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3364, 1.7780, 3.3911, 1.5887, 2.3848, 3.7467, 3.8666, 3.1409], + device='cuda:0'), covar=tensor([0.0942, 0.1581, 0.0311, 0.2158, 0.1014, 0.0232, 0.0491, 0.0551], + device='cuda:0'), in_proj_covar=tensor([0.0293, 0.0317, 0.0282, 0.0312, 0.0308, 0.0265, 0.0419, 0.0300], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 05:56:59,930 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.6356, 1.3985, 4.8834, 1.8242, 4.2952, 4.0391, 4.3755, 4.3011], + device='cuda:0'), covar=tensor([0.0535, 0.5087, 0.0455, 0.4287, 0.1082, 0.1006, 0.0530, 0.0623], + device='cuda:0'), in_proj_covar=tensor([0.0646, 0.0654, 0.0708, 0.0643, 0.0723, 0.0619, 0.0618, 0.0693], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 05:57:01,785 INFO [optim.py:369] (0/4) 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 05:57:04,783 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-184000.pt +2023-02-07 05:57:22,990 INFO [train.py:901] (0/4) Epoch 23, batch 6200, loss[loss=0.2159, simple_loss=0.3017, pruned_loss=0.06507, over 8104.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2868, pruned_loss=0.06072, over 1614461.50 frames. ], batch size: 23, lr: 3.27e-03, grad_scale: 8.0 +2023-02-07 05:57:23,225 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4023, 2.3276, 1.7384, 2.2562, 2.0939, 1.4793, 1.9626, 1.9451], + device='cuda:0'), covar=tensor([0.1485, 0.0415, 0.1280, 0.0562, 0.0670, 0.1636, 0.0944, 0.0943], + device='cuda:0'), in_proj_covar=tensor([0.0351, 0.0231, 0.0332, 0.0306, 0.0298, 0.0338, 0.0341, 0.0315], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 05:57:54,990 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=184068.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:57:59,624 INFO [train.py:901] (0/4) Epoch 23, batch 6250, loss[loss=0.2187, simple_loss=0.2965, pruned_loss=0.07042, over 8481.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2873, pruned_loss=0.0611, over 1617181.39 frames. ], batch size: 25, lr: 3.27e-03, grad_scale: 8.0 +2023-02-07 05:58:03,995 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.6359, 1.9843, 2.1564, 1.8944, 1.4388, 1.9733, 2.3947, 2.1867], + device='cuda:0'), covar=tensor([0.0483, 0.1074, 0.1436, 0.1290, 0.0590, 0.1257, 0.0591, 0.0545], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0152, 0.0189, 0.0159, 0.0100, 0.0162, 0.0111, 0.0143], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:0') +2023-02-07 05:58:06,669 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5925, 2.0672, 3.2764, 1.4199, 2.4574, 2.0321, 1.6778, 2.4264], + device='cuda:0'), covar=tensor([0.1908, 0.2551, 0.0935, 0.4557, 0.1918, 0.3155, 0.2371, 0.2468], + device='cuda:0'), in_proj_covar=tensor([0.0527, 0.0611, 0.0554, 0.0647, 0.0647, 0.0594, 0.0542, 0.0631], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 05:58:11,438 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 05:58:34,478 INFO [train.py:901] (0/4) Epoch 23, batch 6300, loss[loss=0.1712, simple_loss=0.2644, pruned_loss=0.03899, over 7424.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2865, pruned_loss=0.06054, over 1617752.62 frames. ], batch size: 17, lr: 3.27e-03, grad_scale: 8.0 +2023-02-07 05:58:42,583 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.53 vs. limit=5.0 +2023-02-07 05:58:45,746 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=184141.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:59:04,542 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=184166.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:59:10,617 INFO [train.py:901] (0/4) Epoch 23, batch 6350, loss[loss=0.2102, simple_loss=0.2933, pruned_loss=0.06352, over 8564.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2853, pruned_loss=0.06005, over 1615471.96 frames. ], batch size: 39, lr: 3.26e-03, grad_scale: 8.0 +2023-02-07 05:59:25,787 INFO [optim.py:369] (0/4) 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 05:59:32,345 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=184204.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 05:59:46,837 INFO [train.py:901] (0/4) Epoch 23, batch 6400, loss[loss=0.2058, simple_loss=0.2903, pruned_loss=0.06065, over 8471.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2848, pruned_loss=0.05994, over 1613640.53 frames. ], batch size: 25, lr: 3.26e-03, grad_scale: 8.0 +2023-02-07 06:00:22,069 INFO [train.py:901] (0/4) Epoch 23, batch 6450, loss[loss=0.1881, simple_loss=0.2784, pruned_loss=0.04897, over 8328.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2852, pruned_loss=0.06038, over 1614947.03 frames. ], batch size: 26, lr: 3.26e-03, grad_scale: 8.0 +2023-02-07 06:00:31,212 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2486, 1.0854, 2.4363, 1.0531, 1.9465, 1.9424, 2.1826, 2.1798], + device='cuda:0'), covar=tensor([0.1723, 0.4914, 0.1888, 0.5129, 0.2400, 0.1962, 0.1412, 0.1479], + device='cuda:0'), in_proj_covar=tensor([0.0642, 0.0649, 0.0706, 0.0640, 0.0716, 0.0616, 0.0616, 0.0691], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 06:00:37,217 INFO [optim.py:369] (0/4) 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 06:00:54,561 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=184319.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:00:58,476 INFO [train.py:901] (0/4) Epoch 23, batch 6500, loss[loss=0.199, simple_loss=0.2901, pruned_loss=0.05393, over 8031.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2852, pruned_loss=0.06087, over 1611353.31 frames. ], batch size: 22, lr: 3.26e-03, grad_scale: 8.0 +2023-02-07 06:01:01,453 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.62 vs. limit=5.0 +2023-02-07 06:01:13,631 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=184347.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:01:30,730 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=184372.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:01:32,653 INFO [train.py:901] (0/4) Epoch 23, batch 6550, loss[loss=0.2016, simple_loss=0.2805, pruned_loss=0.06131, over 8513.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2868, pruned_loss=0.06141, over 1616254.33 frames. ], batch size: 28, lr: 3.26e-03, grad_scale: 8.0 +2023-02-07 06:01:34,221 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=184377.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:01:48,090 INFO [optim.py:369] (0/4) 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 06:01:51,710 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-07 06:02:00,063 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=184412.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:02:07,252 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6507, 1.8881, 2.0436, 1.3039, 2.1974, 1.4189, 0.6712, 1.9025], + device='cuda:0'), covar=tensor([0.0638, 0.0404, 0.0328, 0.0649, 0.0400, 0.0918, 0.0938, 0.0327], + device='cuda:0'), in_proj_covar=tensor([0.0455, 0.0396, 0.0349, 0.0448, 0.0381, 0.0536, 0.0392, 0.0425], + device='cuda:0'), 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:0') +2023-02-07 06:02:09,832 INFO [train.py:901] (0/4) Epoch 23, batch 6600, loss[loss=0.2611, simple_loss=0.318, pruned_loss=0.1021, over 7164.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2865, pruned_loss=0.0615, over 1610216.01 frames. ], batch size: 71, lr: 3.26e-03, grad_scale: 8.0 +2023-02-07 06:02:09,849 WARNING [train.py:1067] (0/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-07 06:02:20,762 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 +2023-02-07 06:02:42,925 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 +2023-02-07 06:02:45,212 INFO [train.py:901] (0/4) Epoch 23, batch 6650, loss[loss=0.1859, simple_loss=0.2801, pruned_loss=0.04583, over 8456.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.287, pruned_loss=0.06139, over 1608441.67 frames. ], batch size: 27, lr: 3.26e-03, grad_scale: 8.0 +2023-02-07 06:02:52,258 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0223, 1.7062, 3.3011, 1.4968, 2.4784, 3.6092, 3.7307, 3.0449], + device='cuda:0'), covar=tensor([0.1169, 0.1701, 0.0323, 0.2182, 0.0974, 0.0249, 0.0532, 0.0571], + device='cuda:0'), in_proj_covar=tensor([0.0296, 0.0320, 0.0286, 0.0315, 0.0313, 0.0269, 0.0425, 0.0306], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 06:03:00,392 INFO [optim.py:369] (0/4) 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 06:03:06,012 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=184504.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:03:21,176 INFO [train.py:901] (0/4) Epoch 23, batch 6700, loss[loss=0.2266, simple_loss=0.3241, pruned_loss=0.06449, over 8458.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2867, pruned_loss=0.06087, over 1606026.04 frames. ], batch size: 25, lr: 3.26e-03, grad_scale: 8.0 +2023-02-07 06:03:22,771 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=184527.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:03:56,987 INFO [train.py:901] (0/4) Epoch 23, batch 6750, loss[loss=0.1846, simple_loss=0.2705, pruned_loss=0.04941, over 7815.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2873, pruned_loss=0.06125, over 1608919.60 frames. ], batch size: 20, lr: 3.26e-03, grad_scale: 8.0 +2023-02-07 06:03:57,247 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=184575.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:04:11,517 INFO [optim.py:369] (0/4) 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 06:04:14,468 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=184600.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:04:30,935 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-07 06:04:32,267 INFO [train.py:901] (0/4) Epoch 23, batch 6800, loss[loss=0.28, simple_loss=0.3583, pruned_loss=0.1009, over 8596.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2875, pruned_loss=0.06159, over 1612866.41 frames. ], batch size: 31, lr: 3.26e-03, grad_scale: 8.0 +2023-02-07 06:04:55,640 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 +2023-02-07 06:05:08,929 INFO [train.py:901] (0/4) Epoch 23, batch 6850, loss[loss=0.1815, simple_loss=0.2578, pruned_loss=0.05263, over 7435.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2871, pruned_loss=0.06091, over 1614831.38 frames. ], batch size: 17, lr: 3.26e-03, grad_scale: 8.0 +2023-02-07 06:05:19,309 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-07 06:05:23,518 INFO [optim.py:369] (0/4) 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 06:05:40,681 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=184721.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:05:43,468 INFO [train.py:901] (0/4) Epoch 23, batch 6900, loss[loss=0.2303, simple_loss=0.3023, pruned_loss=0.07913, over 8611.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.288, pruned_loss=0.06159, over 1617201.09 frames. ], batch size: 39, lr: 3.26e-03, grad_scale: 8.0 +2023-02-07 06:06:21,325 INFO [train.py:901] (0/4) Epoch 23, batch 6950, loss[loss=0.2095, simple_loss=0.2975, pruned_loss=0.06072, over 8614.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2877, pruned_loss=0.06124, over 1616943.88 frames. ], batch size: 31, lr: 3.26e-03, grad_scale: 8.0 +2023-02-07 06:06:27,153 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=184783.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:06:29,056 WARNING [train.py:1067] (0/4) 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] (0/4) 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 06:06:44,518 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=184808.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:06:56,210 INFO [train.py:901] (0/4) Epoch 23, batch 7000, loss[loss=0.2084, simple_loss=0.2823, pruned_loss=0.06728, over 8508.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2866, pruned_loss=0.06022, over 1621854.07 frames. ], batch size: 26, lr: 3.26e-03, grad_scale: 8.0 +2023-02-07 06:07:03,959 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=184836.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:07:12,026 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 23, batch 7050, loss[loss=0.202, simple_loss=0.2786, pruned_loss=0.06269, over 8030.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2866, pruned_loss=0.0601, over 1620645.51 frames. ], batch size: 22, lr: 3.26e-03, grad_scale: 8.0 +2023-02-07 06:07:38,582 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=184884.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:07:48,062 INFO [optim.py:369] (0/4) 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 06:07:53,399 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.26 vs. limit=5.0 +2023-02-07 06:08:08,171 INFO [train.py:901] (0/4) Epoch 23, batch 7100, loss[loss=0.2139, simple_loss=0.3054, pruned_loss=0.06115, over 8361.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2866, pruned_loss=0.0598, over 1619296.08 frames. ], batch size: 48, lr: 3.26e-03, grad_scale: 8.0 +2023-02-07 06:08:30,154 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=184957.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:08:34,366 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=184963.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:08:42,476 INFO [train.py:901] (0/4) Epoch 23, batch 7150, loss[loss=0.2129, simple_loss=0.2971, pruned_loss=0.06438, over 8337.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.287, pruned_loss=0.06032, over 1619156.87 frames. ], batch size: 26, lr: 3.26e-03, grad_scale: 8.0 +2023-02-07 06:08:53,466 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4870, 1.3495, 1.5875, 1.4362, 1.5458, 1.5081, 1.3799, 0.7882], + device='cuda:0'), covar=tensor([0.3897, 0.3451, 0.1563, 0.2526, 0.1775, 0.2296, 0.1505, 0.3797], + device='cuda:0'), in_proj_covar=tensor([0.0942, 0.0991, 0.0806, 0.0950, 0.0997, 0.0896, 0.0753, 0.0829], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 06:08:58,822 INFO [optim.py:369] (0/4) 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 06:09:20,371 INFO [train.py:901] (0/4) Epoch 23, batch 7200, loss[loss=0.2388, simple_loss=0.3196, pruned_loss=0.07893, over 8101.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2867, pruned_loss=0.06014, over 1619352.59 frames. ], batch size: 23, lr: 3.26e-03, grad_scale: 8.0 +2023-02-07 06:09:29,421 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0559, 1.6536, 1.6880, 1.5492, 0.9128, 1.4848, 1.7651, 1.5353], + device='cuda:0'), covar=tensor([0.0512, 0.1204, 0.1700, 0.1412, 0.0624, 0.1472, 0.0707, 0.0657], + device='cuda:0'), in_proj_covar=tensor([0.0098, 0.0153, 0.0190, 0.0161, 0.0102, 0.0164, 0.0112, 0.0145], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:0') +2023-02-07 06:09:35,027 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9391, 1.5918, 3.5415, 1.6510, 2.4619, 3.9367, 4.0099, 3.4237], + device='cuda:0'), covar=tensor([0.1173, 0.1790, 0.0298, 0.1841, 0.1062, 0.0212, 0.0486, 0.0480], + device='cuda:0'), in_proj_covar=tensor([0.0298, 0.0323, 0.0287, 0.0315, 0.0314, 0.0271, 0.0426, 0.0305], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 06:09:35,757 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.8800, 2.1974, 3.6329, 1.7770, 1.7112, 3.5096, 0.7310, 2.1518], + device='cuda:0'), covar=tensor([0.1328, 0.1263, 0.0219, 0.1890, 0.2640, 0.0306, 0.2225, 0.1426], + device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0195, 0.0129, 0.0219, 0.0268, 0.0136, 0.0168, 0.0191], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 06:09:44,864 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.5420, 1.8082, 2.1945, 1.6949, 0.9091, 1.8638, 1.9619, 1.8553], + device='cuda:0'), covar=tensor([0.0468, 0.1177, 0.1530, 0.1371, 0.0597, 0.1368, 0.0651, 0.0603], + device='cuda:0'), in_proj_covar=tensor([0.0098, 0.0153, 0.0190, 0.0160, 0.0101, 0.0164, 0.0112, 0.0144], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:0') +2023-02-07 06:09:44,903 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.8432, 2.2062, 3.6778, 1.8343, 1.6274, 3.5894, 0.5847, 2.1633], + device='cuda:0'), covar=tensor([0.1524, 0.1345, 0.0249, 0.1942, 0.2826, 0.0363, 0.2488, 0.1525], + device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0196, 0.0129, 0.0220, 0.0268, 0.0136, 0.0169, 0.0191], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 06:09:54,643 INFO [train.py:901] (0/4) Epoch 23, batch 7250, loss[loss=0.215, simple_loss=0.2953, pruned_loss=0.06734, over 8119.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2863, pruned_loss=0.06022, over 1614996.06 frames. ], batch size: 23, lr: 3.26e-03, grad_scale: 8.0 +2023-02-07 06:09:56,879 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=185078.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:10:06,361 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185092.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:10:09,712 INFO [optim.py:369] (0/4) 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 06:10:24,999 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=185117.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:10:30,388 INFO [train.py:901] (0/4) Epoch 23, batch 7300, loss[loss=0.2011, simple_loss=0.2902, pruned_loss=0.05596, over 8571.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2876, pruned_loss=0.06142, over 1616814.76 frames. ], batch size: 39, lr: 3.26e-03, grad_scale: 8.0 +2023-02-07 06:10:45,748 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9137, 1.6574, 1.9801, 1.8153, 1.9093, 1.9547, 1.7649, 0.8393], + device='cuda:0'), covar=tensor([0.5145, 0.4417, 0.2016, 0.3168, 0.2111, 0.2782, 0.1901, 0.4531], + device='cuda:0'), in_proj_covar=tensor([0.0939, 0.0986, 0.0805, 0.0948, 0.0995, 0.0896, 0.0751, 0.0826], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 06:10:46,963 INFO [zipformer.py:1185] (0/4) 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] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.19 vs. limit=5.0 +2023-02-07 06:11:06,510 INFO [train.py:901] (0/4) Epoch 23, batch 7350, loss[loss=0.1863, simple_loss=0.2835, pruned_loss=0.04455, over 8288.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2881, pruned_loss=0.06134, over 1618859.36 frames. ], batch size: 23, lr: 3.26e-03, grad_scale: 8.0 +2023-02-07 06:11:19,726 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-07 06:11:21,063 INFO [optim.py:369] (0/4) 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 06:11:38,654 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7672, 1.7142, 2.4398, 1.5600, 1.2975, 2.4017, 0.5465, 1.4651], + device='cuda:0'), covar=tensor([0.1385, 0.1100, 0.0276, 0.1134, 0.2432, 0.0309, 0.2024, 0.1261], + device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0195, 0.0128, 0.0219, 0.0268, 0.0136, 0.0168, 0.0191], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 06:11:38,659 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185219.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:11:41,240 WARNING [train.py:1067] (0/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-07 06:11:42,634 INFO [train.py:901] (0/4) Epoch 23, batch 7400, loss[loss=0.2345, simple_loss=0.3066, pruned_loss=0.08119, over 8596.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2872, pruned_loss=0.0612, over 1609010.07 frames. ], batch size: 34, lr: 3.26e-03, grad_scale: 16.0 +2023-02-07 06:11:44,779 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=185228.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:11:48,318 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0324, 1.8293, 3.3245, 1.4821, 2.3130, 3.6235, 3.7381, 3.0398], + device='cuda:0'), covar=tensor([0.1163, 0.1602, 0.0330, 0.2150, 0.1073, 0.0236, 0.0565, 0.0563], + device='cuda:0'), in_proj_covar=tensor([0.0297, 0.0322, 0.0287, 0.0315, 0.0314, 0.0269, 0.0424, 0.0305], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 06:11:56,710 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=185244.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:12:04,425 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 +2023-02-07 06:12:18,696 INFO [train.py:901] (0/4) Epoch 23, batch 7450, loss[loss=0.2327, simple_loss=0.3203, pruned_loss=0.07254, over 8466.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2867, pruned_loss=0.0611, over 1614929.83 frames. ], batch size: 29, lr: 3.26e-03, grad_scale: 16.0 +2023-02-07 06:12:18,962 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5783, 1.9385, 2.8123, 1.3962, 2.1586, 1.9212, 1.6512, 2.1809], + device='cuda:0'), covar=tensor([0.1977, 0.2635, 0.1029, 0.4751, 0.2007, 0.3316, 0.2492, 0.2338], + device='cuda:0'), in_proj_covar=tensor([0.0526, 0.0611, 0.0554, 0.0646, 0.0647, 0.0595, 0.0541, 0.0632], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 06:12:21,569 WARNING [train.py:1067] (0/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-07 06:12:33,471 INFO [optim.py:369] (0/4) 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 06:12:37,081 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=185301.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:12:53,821 INFO [train.py:901] (0/4) Epoch 23, batch 7500, loss[loss=0.2079, simple_loss=0.2883, pruned_loss=0.06374, over 8282.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2861, pruned_loss=0.06115, over 1613335.97 frames. ], batch size: 23, lr: 3.25e-03, grad_scale: 16.0 +2023-02-07 06:13:05,840 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 +2023-02-07 06:13:08,254 INFO [zipformer.py:1185] (0/4) 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] (0/4) attn_weights_entropy = tensor([2.4741, 2.8910, 2.1893, 4.0080, 1.5068, 2.0431, 2.4543, 2.7881], + device='cuda:0'), covar=tensor([0.0679, 0.0800, 0.0879, 0.0260, 0.1178, 0.1241, 0.0857, 0.0760], + device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0199, 0.0244, 0.0214, 0.0206, 0.0247, 0.0250, 0.0207], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 06:13:31,746 INFO [train.py:901] (0/4) Epoch 23, batch 7550, loss[loss=0.1968, simple_loss=0.2785, pruned_loss=0.05753, over 7936.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2856, pruned_loss=0.06094, over 1613724.70 frames. ], batch size: 20, lr: 3.25e-03, grad_scale: 16.0 +2023-02-07 06:13:39,223 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 +2023-02-07 06:13:46,072 INFO [optim.py:369] (0/4) 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 06:14:00,499 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=185422.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:14:06,601 INFO [train.py:901] (0/4) Epoch 23, batch 7600, loss[loss=0.1847, simple_loss=0.2543, pruned_loss=0.05759, over 7234.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2847, pruned_loss=0.06038, over 1612556.04 frames. ], batch size: 16, lr: 3.25e-03, grad_scale: 16.0 +2023-02-07 06:14:09,580 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.82 vs. limit=5.0 +2023-02-07 06:14:41,964 INFO [train.py:901] (0/4) Epoch 23, batch 7650, loss[loss=0.2126, simple_loss=0.3051, pruned_loss=0.06005, over 8187.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2851, pruned_loss=0.06097, over 1612109.51 frames. ], batch size: 23, lr: 3.25e-03, grad_scale: 16.0 +2023-02-07 06:14:50,191 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=185486.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:14:54,391 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=185491.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:14:57,818 INFO [optim.py:369] (0/4) 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 06:15:17,441 INFO [train.py:901] (0/4) Epoch 23, batch 7700, loss[loss=0.2108, simple_loss=0.2896, pruned_loss=0.06605, over 8075.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2859, pruned_loss=0.06097, over 1612609.95 frames. ], batch size: 21, lr: 3.25e-03, grad_scale: 16.0 +2023-02-07 06:15:25,724 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=185537.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:15:37,342 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 23, batch 7750, loss[loss=0.2188, simple_loss=0.2923, pruned_loss=0.07264, over 7792.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2853, pruned_loss=0.06067, over 1609109.49 frames. ], batch size: 19, lr: 3.25e-03, grad_scale: 16.0 +2023-02-07 06:16:08,169 INFO [optim.py:369] (0/4) 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 06:16:10,495 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185599.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:16:15,291 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=185606.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:16:28,335 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=185624.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:16:28,831 INFO [train.py:901] (0/4) Epoch 23, batch 7800, loss[loss=0.2014, simple_loss=0.29, pruned_loss=0.05635, over 8475.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2863, pruned_loss=0.06128, over 1607015.41 frames. ], batch size: 25, lr: 3.25e-03, grad_scale: 16.0 +2023-02-07 06:17:01,066 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185672.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:17:02,911 INFO [train.py:901] (0/4) Epoch 23, batch 7850, loss[loss=0.2129, simple_loss=0.2984, pruned_loss=0.06369, over 8555.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2847, pruned_loss=0.06041, over 1607954.73 frames. ], batch size: 49, lr: 3.25e-03, grad_scale: 16.0 +2023-02-07 06:17:17,281 INFO [optim.py:369] (0/4) 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 06:17:18,189 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=185697.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:17:30,684 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.28 vs. limit=5.0 +2023-02-07 06:17:36,388 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 +2023-02-07 06:17:37,222 INFO [train.py:901] (0/4) Epoch 23, batch 7900, loss[loss=0.2125, simple_loss=0.2945, pruned_loss=0.06523, over 8627.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2848, pruned_loss=0.06054, over 1601890.02 frames. ], batch size: 34, lr: 3.25e-03, grad_scale: 16.0 +2023-02-07 06:18:05,868 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5441, 1.7658, 1.9096, 1.2221, 1.9590, 1.4061, 0.4781, 1.7560], + device='cuda:0'), covar=tensor([0.0479, 0.0327, 0.0289, 0.0487, 0.0365, 0.0827, 0.0785, 0.0266], + device='cuda:0'), in_proj_covar=tensor([0.0461, 0.0401, 0.0354, 0.0454, 0.0386, 0.0544, 0.0398, 0.0430], + device='cuda:0'), out_proj_covar=tensor([1.2310e-04, 1.0477e-04, 9.3033e-05, 1.1932e-04, 1.0143e-04, 1.5319e-04, + 1.0712e-04, 1.1351e-04], device='cuda:0') +2023-02-07 06:18:11,090 INFO [train.py:901] (0/4) Epoch 23, batch 7950, loss[loss=0.1748, simple_loss=0.2547, pruned_loss=0.04749, over 7812.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2843, pruned_loss=0.06007, over 1603133.75 frames. ], batch size: 20, lr: 3.25e-03, grad_scale: 16.0 +2023-02-07 06:18:12,591 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=185777.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:18:23,319 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 06:18:39,370 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=185817.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 06:18:40,117 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=185818.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:18:44,667 INFO [train.py:901] (0/4) Epoch 23, batch 8000, loss[loss=0.2622, simple_loss=0.3337, pruned_loss=0.0953, over 8670.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2862, pruned_loss=0.06109, over 1605593.96 frames. ], batch size: 39, lr: 3.25e-03, grad_scale: 16.0 +2023-02-07 06:18:48,036 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=185830.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:19:09,864 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185862.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:19:18,260 INFO [train.py:901] (0/4) Epoch 23, batch 8050, loss[loss=0.1792, simple_loss=0.2538, pruned_loss=0.05233, over 7529.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2841, pruned_loss=0.06082, over 1589089.33 frames. ], batch size: 18, lr: 3.25e-03, grad_scale: 16.0 +2023-02-07 06:19:26,746 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=185887.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:19:32,784 INFO [optim.py:369] (0/4) 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 06:19:41,826 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/epoch-23.pt +2023-02-07 06:19:53,796 WARNING [train.py:1067] (0/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-07 06:19:57,770 INFO [train.py:901] (0/4) Epoch 24, batch 0, loss[loss=0.1683, simple_loss=0.2475, pruned_loss=0.04454, over 7238.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2475, pruned_loss=0.04454, over 7238.00 frames. ], batch size: 16, lr: 3.18e-03, grad_scale: 16.0 +2023-02-07 06:19:57,771 INFO [train.py:926] (0/4) Computing validation loss +2023-02-07 06:20:01,566 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5874, 1.3106, 1.5603, 1.2867, 0.9029, 1.3291, 1.5476, 1.2196], + device='cuda:0'), covar=tensor([0.0647, 0.1395, 0.1778, 0.1557, 0.0639, 0.1603, 0.0721, 0.0735], + device='cuda:0'), in_proj_covar=tensor([0.0098, 0.0153, 0.0190, 0.0160, 0.0101, 0.0164, 0.0112, 0.0144], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:0') +2023-02-07 06:20:09,065 INFO [train.py:935] (0/4) Epoch 24, validation: loss=0.1731, simple_loss=0.2733, pruned_loss=0.03644, over 944034.00 frames. +2023-02-07 06:20:09,066 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-07 06:20:22,818 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7259, 2.0177, 3.3096, 1.5522, 2.6104, 2.1185, 1.7762, 2.7111], + device='cuda:0'), covar=tensor([0.1995, 0.2815, 0.0795, 0.4463, 0.1811, 0.3084, 0.2388, 0.2125], + device='cuda:0'), in_proj_covar=tensor([0.0525, 0.0610, 0.0551, 0.0645, 0.0647, 0.0594, 0.0542, 0.0631], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 06:20:23,944 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-07 06:20:35,528 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=185945.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:20:44,032 INFO [train.py:901] (0/4) Epoch 24, batch 50, loss[loss=0.1775, simple_loss=0.264, pruned_loss=0.04556, over 8519.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2864, pruned_loss=0.06094, over 365468.98 frames. ], batch size: 26, lr: 3.18e-03, grad_scale: 16.0 +2023-02-07 06:20:57,554 WARNING [train.py:1067] (0/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-07 06:21:11,394 INFO [optim.py:369] (0/4) 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 06:21:14,405 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-186000.pt +2023-02-07 06:21:16,891 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2951, 2.8774, 2.1488, 4.0976, 1.6841, 1.7889, 2.3607, 2.9044], + device='cuda:0'), covar=tensor([0.0747, 0.0702, 0.0888, 0.0291, 0.1046, 0.1425, 0.1038, 0.0760], + device='cuda:0'), in_proj_covar=tensor([0.0230, 0.0197, 0.0243, 0.0214, 0.0206, 0.0247, 0.0250, 0.0206], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 06:21:20,546 INFO [train.py:901] (0/4) Epoch 24, batch 100, loss[loss=0.2291, simple_loss=0.3067, pruned_loss=0.07571, over 8478.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2891, pruned_loss=0.06103, over 648178.04 frames. ], batch size: 28, lr: 3.18e-03, grad_scale: 16.0 +2023-02-07 06:21:22,589 WARNING [train.py:1067] (0/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-07 06:21:44,089 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-02-07 06:21:56,095 INFO [train.py:901] (0/4) Epoch 24, batch 150, loss[loss=0.1809, simple_loss=0.2663, pruned_loss=0.04779, over 8088.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2875, pruned_loss=0.061, over 860587.84 frames. ], batch size: 21, lr: 3.18e-03, grad_scale: 16.0 +2023-02-07 06:22:21,896 INFO [optim.py:369] (0/4) 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 06:22:30,263 INFO [train.py:901] (0/4) Epoch 24, batch 200, loss[loss=0.174, simple_loss=0.2526, pruned_loss=0.04766, over 7814.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2859, pruned_loss=0.0601, over 1025636.82 frames. ], batch size: 20, lr: 3.18e-03, grad_scale: 16.0 +2023-02-07 06:22:38,134 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186118.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:22:40,016 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=186121.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:22:43,726 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.88 vs. limit=5.0 +2023-02-07 06:23:05,568 INFO [train.py:901] (0/4) Epoch 24, batch 250, loss[loss=0.2241, simple_loss=0.312, pruned_loss=0.06815, over 8500.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2859, pruned_loss=0.06013, over 1155636.54 frames. ], batch size: 28, lr: 3.18e-03, grad_scale: 16.0 +2023-02-07 06:23:07,692 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=186161.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 06:23:16,539 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-07 06:23:18,798 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186176.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:23:25,610 WARNING [train.py:1067] (0/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-07 06:23:32,338 INFO [optim.py:369] (0/4) 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 06:23:36,036 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186201.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:23:40,507 INFO [train.py:901] (0/4) Epoch 24, batch 300, loss[loss=0.1845, simple_loss=0.2523, pruned_loss=0.05828, over 7424.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2889, pruned_loss=0.06225, over 1258570.14 frames. ], batch size: 17, lr: 3.18e-03, grad_scale: 16.0 +2023-02-07 06:23:49,145 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 +2023-02-07 06:23:50,381 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.52 vs. limit=2.0 +2023-02-07 06:23:52,985 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186226.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:24:00,507 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186236.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:24:15,085 INFO [train.py:901] (0/4) Epoch 24, batch 350, loss[loss=0.1962, simple_loss=0.2773, pruned_loss=0.05761, over 7659.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.289, pruned_loss=0.06223, over 1335930.90 frames. ], batch size: 19, lr: 3.18e-03, grad_scale: 16.0 +2023-02-07 06:24:28,126 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186276.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 06:24:28,699 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186277.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:24:42,189 INFO [optim.py:369] (0/4) 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 06:24:47,223 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-02-07 06:24:50,342 INFO [train.py:901] (0/4) Epoch 24, batch 400, loss[loss=0.1909, simple_loss=0.2844, pruned_loss=0.04865, over 8280.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2894, pruned_loss=0.06227, over 1396652.91 frames. ], batch size: 23, lr: 3.18e-03, grad_scale: 16.0 +2023-02-07 06:25:26,090 INFO [train.py:901] (0/4) Epoch 24, batch 450, loss[loss=0.1804, simple_loss=0.2534, pruned_loss=0.05372, over 7776.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2882, pruned_loss=0.06205, over 1441974.09 frames. ], batch size: 19, lr: 3.18e-03, grad_scale: 16.0 +2023-02-07 06:25:52,934 INFO [optim.py:369] (0/4) 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 06:26:02,009 INFO [train.py:901] (0/4) Epoch 24, batch 500, loss[loss=0.1818, simple_loss=0.2607, pruned_loss=0.05145, over 7975.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.288, pruned_loss=0.06169, over 1485208.97 frames. ], batch size: 21, lr: 3.18e-03, grad_scale: 16.0 +2023-02-07 06:26:11,160 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5277, 2.7273, 1.9484, 2.3099, 2.0835, 1.7451, 2.1648, 2.2577], + device='cuda:0'), covar=tensor([0.1516, 0.0389, 0.1253, 0.0648, 0.0782, 0.1457, 0.0982, 0.0919], + device='cuda:0'), in_proj_covar=tensor([0.0358, 0.0236, 0.0338, 0.0312, 0.0304, 0.0343, 0.0350, 0.0320], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 06:26:23,333 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186439.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:26:37,814 INFO [train.py:901] (0/4) Epoch 24, batch 550, loss[loss=0.1836, simple_loss=0.2688, pruned_loss=0.04916, over 7806.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2858, pruned_loss=0.0604, over 1508615.92 frames. ], batch size: 20, lr: 3.18e-03, grad_scale: 16.0 +2023-02-07 06:26:40,765 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=186462.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:26:42,861 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6065, 1.6228, 2.2927, 1.5040, 1.1767, 2.2309, 0.5105, 1.3024], + device='cuda:0'), covar=tensor([0.1848, 0.1423, 0.0361, 0.1194, 0.2735, 0.0389, 0.1780, 0.1447], + device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0196, 0.0129, 0.0217, 0.0268, 0.0135, 0.0168, 0.0190], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 06:27:01,103 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186492.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:27:01,700 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 06:27:05,957 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1575, 1.8980, 2.4575, 2.0184, 2.4281, 2.1887, 1.9639, 1.1746], + device='cuda:0'), covar=tensor([0.5784, 0.5148, 0.2094, 0.4019, 0.2666, 0.3313, 0.2051, 0.5617], + device='cuda:0'), in_proj_covar=tensor([0.0943, 0.0986, 0.0806, 0.0950, 0.0994, 0.0896, 0.0752, 0.0824], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 06:27:12,584 INFO [train.py:901] (0/4) Epoch 24, batch 600, loss[loss=0.1934, simple_loss=0.2845, pruned_loss=0.05114, over 8246.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2852, pruned_loss=0.06003, over 1530797.23 frames. ], batch size: 24, lr: 3.17e-03, grad_scale: 16.0 +2023-02-07 06:27:19,616 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186517.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:27:19,986 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 +2023-02-07 06:27:21,524 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=186520.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:27:24,467 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-07 06:27:26,213 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-07 06:27:29,777 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186532.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 06:27:31,735 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186535.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:27:31,829 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7781, 2.3244, 4.0848, 1.5502, 3.0700, 2.3058, 1.9815, 2.9387], + device='cuda:0'), covar=tensor([0.1935, 0.2629, 0.0927, 0.4643, 0.1808, 0.3180, 0.2259, 0.2526], + device='cuda:0'), in_proj_covar=tensor([0.0523, 0.0610, 0.0552, 0.0645, 0.0648, 0.0593, 0.0541, 0.0631], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 06:27:42,704 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 +2023-02-07 06:27:46,456 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186557.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:27:46,521 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186557.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 06:27:47,019 INFO [train.py:901] (0/4) Epoch 24, batch 650, loss[loss=0.2342, simple_loss=0.32, pruned_loss=0.0742, over 8444.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2865, pruned_loss=0.06048, over 1551985.89 frames. ], batch size: 29, lr: 3.17e-03, grad_scale: 8.0 +2023-02-07 06:27:49,556 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 +2023-02-07 06:28:01,232 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186577.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:28:07,489 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186585.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:28:15,660 INFO [optim.py:369] (0/4) 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 06:28:20,740 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186604.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:28:23,403 INFO [train.py:901] (0/4) Epoch 24, batch 700, loss[loss=0.2372, simple_loss=0.3146, pruned_loss=0.07987, over 7082.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2864, pruned_loss=0.06039, over 1565456.66 frames. ], batch size: 71, lr: 3.17e-03, grad_scale: 8.0 +2023-02-07 06:28:33,205 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=186621.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:28:43,729 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186635.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:28:59,664 INFO [train.py:901] (0/4) Epoch 24, batch 750, loss[loss=0.1896, simple_loss=0.268, pruned_loss=0.05557, over 8151.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.286, pruned_loss=0.06014, over 1576374.96 frames. ], batch size: 22, lr: 3.17e-03, grad_scale: 8.0 +2023-02-07 06:29:11,917 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-07 06:29:21,535 WARNING [train.py:1067] (0/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-07 06:29:27,091 INFO [optim.py:369] (0/4) 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 06:29:35,751 INFO [train.py:901] (0/4) Epoch 24, batch 800, loss[loss=0.2146, simple_loss=0.3048, pruned_loss=0.0622, over 8328.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2855, pruned_loss=0.0599, over 1583887.58 frames. ], batch size: 25, lr: 3.17e-03, grad_scale: 8.0 +2023-02-07 06:29:48,922 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186727.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:29:48,941 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9619, 1.4908, 3.3788, 1.5760, 2.3499, 3.7079, 3.8236, 3.1696], + device='cuda:0'), covar=tensor([0.1189, 0.1879, 0.0335, 0.2009, 0.1164, 0.0231, 0.0517, 0.0532], + device='cuda:0'), in_proj_covar=tensor([0.0296, 0.0321, 0.0285, 0.0314, 0.0312, 0.0268, 0.0424, 0.0302], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 06:29:56,169 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186736.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:29:56,181 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6541, 1.9065, 1.6754, 2.2867, 0.9877, 1.5005, 1.7374, 1.8680], + device='cuda:0'), covar=tensor([0.0781, 0.0696, 0.0891, 0.0426, 0.1092, 0.1300, 0.0715, 0.0677], + device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0195, 0.0242, 0.0213, 0.0204, 0.0244, 0.0248, 0.0204], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 06:30:11,911 INFO [train.py:901] (0/4) Epoch 24, batch 850, loss[loss=0.1757, simple_loss=0.2613, pruned_loss=0.04503, over 7929.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2855, pruned_loss=0.0602, over 1589253.26 frames. ], batch size: 20, lr: 3.17e-03, grad_scale: 8.0 +2023-02-07 06:30:29,490 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=186783.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:30:39,059 INFO [optim.py:369] (0/4) 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 06:30:43,835 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.60 vs. limit=5.0 +2023-02-07 06:30:47,651 INFO [train.py:901] (0/4) Epoch 24, batch 900, loss[loss=0.2075, simple_loss=0.2989, pruned_loss=0.05811, over 7974.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2846, pruned_loss=0.0594, over 1592899.91 frames. ], batch size: 21, lr: 3.17e-03, grad_scale: 8.0 +2023-02-07 06:31:05,955 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186833.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:31:08,567 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=186837.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:31:24,270 INFO [train.py:901] (0/4) Epoch 24, batch 950, loss[loss=0.1881, simple_loss=0.2646, pruned_loss=0.05577, over 7927.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2859, pruned_loss=0.06019, over 1600518.61 frames. ], batch size: 20, lr: 3.17e-03, grad_scale: 8.0 +2023-02-07 06:31:24,442 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186858.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:31:24,504 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186858.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:31:39,988 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=186879.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:31:43,459 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-07 06:31:48,329 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186891.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:31:52,257 INFO [optim.py:369] (0/4) 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 06:31:53,149 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186898.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:31:55,086 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 24, batch 1000, loss[loss=0.2062, simple_loss=0.302, pruned_loss=0.05522, over 8454.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2862, pruned_loss=0.06009, over 1608489.41 frames. ], batch size: 25, lr: 3.17e-03, grad_scale: 8.0 +2023-02-07 06:32:05,645 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186916.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:32:15,385 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=186929.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:32:15,459 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186929.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:32:20,491 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-07 06:32:29,429 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=186948.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:32:32,167 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186952.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:32:33,360 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-07 06:32:36,101 INFO [train.py:901] (0/4) Epoch 24, batch 1050, loss[loss=0.196, simple_loss=0.2782, pruned_loss=0.05686, over 7549.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2854, pruned_loss=0.0594, over 1612828.32 frames. ], batch size: 18, lr: 3.17e-03, grad_scale: 8.0 +2023-02-07 06:32:58,701 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.7448, 2.9748, 2.5428, 4.0567, 1.5275, 2.2676, 2.5264, 2.9363], + device='cuda:0'), covar=tensor([0.0607, 0.0718, 0.0671, 0.0227, 0.1068, 0.1122, 0.0893, 0.0780], + device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0195, 0.0242, 0.0213, 0.0204, 0.0244, 0.0248, 0.0204], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 06:33:01,536 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186992.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:33:02,855 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186994.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:33:04,739 INFO [optim.py:369] (0/4) 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 06:33:12,439 INFO [train.py:901] (0/4) Epoch 24, batch 1100, loss[loss=0.1873, simple_loss=0.2777, pruned_loss=0.04841, over 8362.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2848, pruned_loss=0.05903, over 1610601.04 frames. ], batch size: 24, lr: 3.17e-03, grad_scale: 8.0 +2023-02-07 06:33:12,711 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4696, 1.7154, 2.6478, 1.3585, 1.9302, 1.8142, 1.4921, 1.8767], + device='cuda:0'), covar=tensor([0.2018, 0.2721, 0.0917, 0.4831, 0.2043, 0.3418, 0.2617, 0.2390], + device='cuda:0'), in_proj_covar=tensor([0.0527, 0.0615, 0.0556, 0.0650, 0.0653, 0.0600, 0.0544, 0.0634], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 06:33:18,325 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187016.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:33:19,056 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187017.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:33:37,504 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187043.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:33:38,266 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187044.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:33:45,685 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-07 06:33:47,119 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187056.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:33:48,344 INFO [train.py:901] (0/4) Epoch 24, batch 1150, loss[loss=0.1792, simple_loss=0.2588, pruned_loss=0.04983, over 8236.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2863, pruned_loss=0.05976, over 1615445.66 frames. ], batch size: 22, lr: 3.17e-03, grad_scale: 8.0 +2023-02-07 06:33:49,911 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0225, 1.6933, 3.4316, 1.7630, 2.5724, 3.7391, 3.8458, 3.2641], + device='cuda:0'), covar=tensor([0.1185, 0.1756, 0.0365, 0.1880, 0.1047, 0.0226, 0.0566, 0.0481], + device='cuda:0'), in_proj_covar=tensor([0.0298, 0.0324, 0.0288, 0.0317, 0.0315, 0.0270, 0.0427, 0.0304], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 06:33:52,026 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187063.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:33:57,506 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=187071.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:34:02,539 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6051, 2.5569, 1.8178, 2.3914, 2.0831, 1.6101, 2.1317, 2.2155], + device='cuda:0'), covar=tensor([0.1511, 0.0467, 0.1325, 0.0670, 0.0783, 0.1627, 0.1014, 0.0993], + device='cuda:0'), in_proj_covar=tensor([0.0355, 0.0234, 0.0336, 0.0309, 0.0301, 0.0341, 0.0346, 0.0318], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 06:34:16,153 INFO [optim.py:369] (0/4) 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 06:34:24,624 INFO [train.py:901] (0/4) Epoch 24, batch 1200, loss[loss=0.1575, simple_loss=0.2428, pruned_loss=0.03608, over 7711.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.286, pruned_loss=0.05926, over 1618741.41 frames. ], batch size: 18, lr: 3.17e-03, grad_scale: 8.0 +2023-02-07 06:34:56,620 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 24, batch 1250, loss[loss=0.1477, simple_loss=0.2267, pruned_loss=0.03431, over 7685.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2861, pruned_loss=0.05922, over 1621847.62 frames. ], batch size: 18, lr: 3.17e-03, grad_scale: 4.0 +2023-02-07 06:35:15,142 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187179.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:35:19,761 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187186.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:35:27,952 INFO [optim.py:369] (0/4) 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 06:35:30,862 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=187202.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:35:35,739 INFO [train.py:901] (0/4) Epoch 24, batch 1300, loss[loss=0.1938, simple_loss=0.2739, pruned_loss=0.05679, over 8068.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2857, pruned_loss=0.05928, over 1623707.47 frames. ], batch size: 21, lr: 3.17e-03, grad_scale: 4.0 +2023-02-07 06:35:35,959 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187208.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:35:51,427 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.99 vs. limit=5.0 +2023-02-07 06:35:53,888 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187233.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:36:05,767 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187250.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:36:10,937 INFO [train.py:901] (0/4) Epoch 24, batch 1350, loss[loss=0.2253, simple_loss=0.304, pruned_loss=0.07328, over 8328.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2858, pruned_loss=0.05974, over 1618304.89 frames. ], batch size: 25, lr: 3.17e-03, grad_scale: 4.0 +2023-02-07 06:36:20,754 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187272.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:36:21,291 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=187273.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:36:22,811 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187275.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:36:39,303 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187297.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:36:39,749 INFO [optim.py:369] (0/4) 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 06:36:41,380 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 24, batch 1400, loss[loss=0.1687, simple_loss=0.2568, pruned_loss=0.04029, over 7640.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2861, pruned_loss=0.0601, over 1617064.19 frames. ], batch size: 19, lr: 3.17e-03, grad_scale: 4.0 +2023-02-07 06:36:52,964 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187319.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:36:59,256 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187325.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:37:03,361 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4038, 1.5879, 2.0844, 1.3104, 1.4153, 1.6933, 1.4630, 1.3476], + device='cuda:0'), covar=tensor([0.2020, 0.2493, 0.0959, 0.4570, 0.2041, 0.3346, 0.2453, 0.2341], + device='cuda:0'), in_proj_covar=tensor([0.0529, 0.0617, 0.0558, 0.0652, 0.0653, 0.0600, 0.0545, 0.0636], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 06:37:05,349 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4265, 2.1264, 1.6223, 1.9597, 1.8114, 1.4341, 1.7706, 1.7042], + device='cuda:0'), covar=tensor([0.1129, 0.0397, 0.1244, 0.0525, 0.0699, 0.1456, 0.0880, 0.0810], + device='cuda:0'), in_proj_covar=tensor([0.0353, 0.0233, 0.0335, 0.0308, 0.0300, 0.0338, 0.0346, 0.0316], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 06:37:12,868 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187344.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:37:16,721 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 +2023-02-07 06:37:21,713 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-07 06:37:22,425 INFO [train.py:901] (0/4) Epoch 24, batch 1450, loss[loss=0.1632, simple_loss=0.249, pruned_loss=0.03867, over 8076.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2863, pruned_loss=0.06054, over 1615577.11 frames. ], batch size: 21, lr: 3.17e-03, grad_scale: 4.0 +2023-02-07 06:37:42,313 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=187387.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:37:43,076 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) 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 06:37:51,135 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=187400.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:37:56,982 INFO [train.py:901] (0/4) Epoch 24, batch 1500, loss[loss=0.1974, simple_loss=0.2979, pruned_loss=0.04852, over 8461.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2861, pruned_loss=0.0605, over 1612536.24 frames. ], batch size: 25, lr: 3.17e-03, grad_scale: 4.0 +2023-02-07 06:38:22,153 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 24, batch 1550, loss[loss=0.1919, simple_loss=0.2765, pruned_loss=0.05358, over 8047.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2855, pruned_loss=0.06066, over 1608341.56 frames. ], batch size: 22, lr: 3.17e-03, grad_scale: 4.0 +2023-02-07 06:38:39,577 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187467.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:38:59,896 INFO [optim.py:369] (0/4) 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 06:39:02,834 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 24, batch 1600, loss[loss=0.2129, simple_loss=0.3066, pruned_loss=0.05963, over 8355.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2849, pruned_loss=0.06055, over 1608318.54 frames. ], batch size: 26, lr: 3.17e-03, grad_scale: 8.0 +2023-02-07 06:39:11,504 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187530.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:39:27,620 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187537.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:39:30,378 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187541.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:39:41,433 INFO [train.py:901] (0/4) Epoch 24, batch 1650, loss[loss=0.17, simple_loss=0.2558, pruned_loss=0.04209, over 7246.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2848, pruned_loss=0.06058, over 1610677.10 frames. ], batch size: 16, lr: 3.17e-03, grad_scale: 8.0 +2023-02-07 06:39:52,499 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 06:40:09,872 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187598.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:40:10,442 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9315, 1.6601, 3.2240, 1.5379, 2.3003, 3.5134, 3.6471, 3.0991], + device='cuda:0'), covar=tensor([0.1199, 0.1619, 0.0394, 0.2186, 0.1149, 0.0243, 0.0593, 0.0508], + device='cuda:0'), in_proj_covar=tensor([0.0295, 0.0319, 0.0283, 0.0312, 0.0311, 0.0267, 0.0422, 0.0301], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 06:40:16,425 INFO [train.py:901] (0/4) Epoch 24, batch 1700, loss[loss=0.1665, simple_loss=0.2614, pruned_loss=0.03578, over 8292.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.285, pruned_loss=0.06044, over 1614941.54 frames. ], batch size: 23, lr: 3.17e-03, grad_scale: 8.0 +2023-02-07 06:40:40,843 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187644.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:40:50,906 INFO [train.py:901] (0/4) Epoch 24, batch 1750, loss[loss=0.2312, simple_loss=0.3118, pruned_loss=0.07533, over 8388.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2844, pruned_loss=0.06052, over 1611598.04 frames. ], batch size: 48, lr: 3.17e-03, grad_scale: 8.0 +2023-02-07 06:40:58,599 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187669.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:41:18,561 INFO [optim.py:369] (0/4) 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] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 +2023-02-07 06:41:26,178 INFO [train.py:901] (0/4) Epoch 24, batch 1800, loss[loss=0.1502, simple_loss=0.2386, pruned_loss=0.0309, over 8031.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2836, pruned_loss=0.0597, over 1611469.27 frames. ], batch size: 22, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:41:43,267 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 24, batch 1850, loss[loss=0.2322, simple_loss=0.3131, pruned_loss=0.07565, over 8238.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2848, pruned_loss=0.06045, over 1613799.42 frames. ], batch size: 22, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:41:59,465 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187758.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:42:09,287 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187771.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:42:14,870 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6163, 1.5177, 2.8801, 1.3719, 2.2847, 3.0770, 3.2723, 2.6661], + device='cuda:0'), covar=tensor([0.1203, 0.1535, 0.0346, 0.2148, 0.0792, 0.0303, 0.0617, 0.0564], + device='cuda:0'), in_proj_covar=tensor([0.0295, 0.0319, 0.0284, 0.0312, 0.0312, 0.0267, 0.0423, 0.0301], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 06:42:18,445 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187783.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:42:27,398 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 06:42:36,381 INFO [train.py:901] (0/4) Epoch 24, batch 1900, loss[loss=0.2043, simple_loss=0.291, pruned_loss=0.05882, over 8592.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2848, pruned_loss=0.06033, over 1613549.74 frames. ], batch size: 34, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:42:42,643 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.8665, 5.9233, 5.1572, 2.7406, 5.3010, 5.6767, 5.4598, 5.4917], + device='cuda:0'), covar=tensor([0.0521, 0.0422, 0.0859, 0.4147, 0.0755, 0.0850, 0.1128, 0.0530], + device='cuda:0'), in_proj_covar=tensor([0.0529, 0.0446, 0.0435, 0.0543, 0.0435, 0.0448, 0.0428, 0.0391], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 06:42:49,918 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.49 vs. limit=2.0 +2023-02-07 06:43:02,032 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 +2023-02-07 06:43:05,617 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-07 06:43:05,793 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187849.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:43:11,793 INFO [train.py:901] (0/4) Epoch 24, batch 1950, loss[loss=0.2007, simple_loss=0.2813, pruned_loss=0.06002, over 8125.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2847, pruned_loss=0.05991, over 1616230.15 frames. ], batch size: 22, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:43:18,405 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-07 06:43:22,597 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=187874.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:43:27,982 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=187881.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:43:30,722 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=187885.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:43:39,203 INFO [optim.py:369] (0/4) 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 06:43:39,247 WARNING [train.py:1067] (0/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-07 06:43:42,654 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5420, 1.8172, 1.9622, 1.1909, 2.0831, 1.3176, 0.6958, 1.8163], + device='cuda:0'), covar=tensor([0.0825, 0.0494, 0.0380, 0.0874, 0.0538, 0.1399, 0.1097, 0.0421], + device='cuda:0'), in_proj_covar=tensor([0.0458, 0.0401, 0.0355, 0.0451, 0.0385, 0.0541, 0.0394, 0.0426], + device='cuda:0'), 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:0') +2023-02-07 06:43:46,280 INFO [train.py:901] (0/4) Epoch 24, batch 2000, loss[loss=0.2007, simple_loss=0.2858, pruned_loss=0.05782, over 7654.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2847, pruned_loss=0.0601, over 1614222.92 frames. ], batch size: 19, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:44:21,546 INFO [train.py:901] (0/4) Epoch 24, batch 2050, loss[loss=0.2184, simple_loss=0.2952, pruned_loss=0.07082, over 8141.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2861, pruned_loss=0.06063, over 1618284.27 frames. ], batch size: 22, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:44:27,168 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2062, 2.0434, 2.6408, 2.2376, 2.6196, 2.2587, 2.0398, 1.4904], + device='cuda:0'), covar=tensor([0.5686, 0.5080, 0.2099, 0.3878, 0.2658, 0.3274, 0.2080, 0.5384], + device='cuda:0'), in_proj_covar=tensor([0.0951, 0.0999, 0.0816, 0.0964, 0.1001, 0.0907, 0.0761, 0.0833], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 06:44:34,897 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2310, 1.9998, 2.5708, 2.1573, 2.5663, 2.2841, 2.1009, 1.3878], + device='cuda:0'), covar=tensor([0.5571, 0.4895, 0.2007, 0.3907, 0.2437, 0.3302, 0.2146, 0.5202], + device='cuda:0'), in_proj_covar=tensor([0.0950, 0.0998, 0.0815, 0.0963, 0.1000, 0.0907, 0.0761, 0.0832], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 06:44:35,436 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.2479, 3.1561, 2.9688, 1.6738, 2.8740, 2.9213, 2.8596, 2.8335], + device='cuda:0'), covar=tensor([0.1164, 0.0847, 0.1331, 0.4293, 0.1115, 0.1202, 0.1653, 0.1008], + device='cuda:0'), in_proj_covar=tensor([0.0527, 0.0445, 0.0434, 0.0541, 0.0433, 0.0447, 0.0425, 0.0387], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 06:44:42,290 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187989.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:44:46,897 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 06:44:50,345 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-188000.pt +2023-02-07 06:44:51,434 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=188000.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:44:56,704 INFO [train.py:901] (0/4) Epoch 24, batch 2100, loss[loss=0.1536, simple_loss=0.2357, pruned_loss=0.03576, over 7701.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2856, pruned_loss=0.05991, over 1620768.85 frames. ], batch size: 18, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:45:04,573 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.88 vs. limit=5.0 +2023-02-07 06:45:32,168 INFO [train.py:901] (0/4) Epoch 24, batch 2150, loss[loss=0.1554, simple_loss=0.2361, pruned_loss=0.03739, over 7803.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2853, pruned_loss=0.06005, over 1619204.56 frames. ], batch size: 19, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:45:45,161 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3938, 1.6232, 1.6606, 1.1626, 1.6613, 1.3205, 0.3339, 1.6046], + device='cuda:0'), covar=tensor([0.0465, 0.0355, 0.0316, 0.0483, 0.0468, 0.0936, 0.0878, 0.0263], + device='cuda:0'), in_proj_covar=tensor([0.0456, 0.0398, 0.0353, 0.0448, 0.0383, 0.0537, 0.0392, 0.0424], + device='cuda:0'), out_proj_covar=tensor([1.2176e-04, 1.0405e-04, 9.2593e-05, 1.1775e-04, 1.0065e-04, 1.5101e-04, + 1.0535e-04, 1.1185e-04], device='cuda:0') +2023-02-07 06:45:58,768 INFO [optim.py:369] (0/4) 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 06:46:03,905 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=188105.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:46:05,707 INFO [train.py:901] (0/4) Epoch 24, batch 2200, loss[loss=0.2396, simple_loss=0.3227, pruned_loss=0.07828, over 8537.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2852, pruned_loss=0.05962, over 1620585.08 frames. ], batch size: 49, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:46:21,941 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=188130.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:46:41,365 INFO [train.py:901] (0/4) Epoch 24, batch 2250, loss[loss=0.2091, simple_loss=0.3078, pruned_loss=0.05518, over 8257.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2839, pruned_loss=0.05876, over 1616798.78 frames. ], batch size: 24, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:46:46,285 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-07 06:46:58,996 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.1356, 4.0934, 3.6723, 2.0307, 3.6119, 3.7294, 3.6738, 3.5775], + device='cuda:0'), covar=tensor([0.0858, 0.0656, 0.1178, 0.4475, 0.0951, 0.1151, 0.1366, 0.0836], + device='cuda:0'), in_proj_covar=tensor([0.0529, 0.0445, 0.0436, 0.0543, 0.0433, 0.0447, 0.0427, 0.0389], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 06:47:09,411 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 24, batch 2300, loss[loss=0.1877, simple_loss=0.2635, pruned_loss=0.05596, over 5576.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2832, pruned_loss=0.0587, over 1613384.90 frames. ], batch size: 12, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:47:42,352 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=188245.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:47:47,116 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=188252.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:47:50,571 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=188256.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:47:51,705 INFO [train.py:901] (0/4) Epoch 24, batch 2350, loss[loss=0.2771, simple_loss=0.3402, pruned_loss=0.107, over 6996.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.284, pruned_loss=0.05895, over 1616808.52 frames. ], batch size: 71, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:48:00,163 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=188270.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:48:05,490 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=188277.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:48:08,173 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=188281.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:48:20,113 INFO [optim.py:369] (0/4) 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 06:48:21,112 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2743, 2.0828, 2.7035, 2.3308, 2.7101, 2.3401, 2.1491, 1.6190], + device='cuda:0'), covar=tensor([0.5551, 0.5149, 0.1875, 0.3398, 0.2413, 0.3100, 0.2001, 0.5323], + device='cuda:0'), in_proj_covar=tensor([0.0948, 0.0998, 0.0816, 0.0960, 0.1000, 0.0906, 0.0760, 0.0834], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 06:48:22,415 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=188301.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:48:27,108 INFO [train.py:901] (0/4) Epoch 24, batch 2400, loss[loss=0.2237, simple_loss=0.2928, pruned_loss=0.0773, over 7782.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2845, pruned_loss=0.05926, over 1612162.78 frames. ], batch size: 19, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:48:55,189 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 +2023-02-07 06:49:02,253 INFO [train.py:901] (0/4) Epoch 24, batch 2450, loss[loss=0.2043, simple_loss=0.2764, pruned_loss=0.06612, over 8090.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2848, pruned_loss=0.0592, over 1612054.44 frames. ], batch size: 21, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:49:20,761 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5374, 1.7980, 1.8543, 1.1704, 1.9343, 1.4367, 0.4484, 1.7956], + device='cuda:0'), covar=tensor([0.0548, 0.0368, 0.0301, 0.0599, 0.0399, 0.0921, 0.0943, 0.0278], + device='cuda:0'), in_proj_covar=tensor([0.0458, 0.0399, 0.0355, 0.0451, 0.0384, 0.0540, 0.0395, 0.0426], + device='cuda:0'), out_proj_covar=tensor([1.2227e-04, 1.0430e-04, 9.3155e-05, 1.1845e-04, 1.0094e-04, 1.5171e-04, + 1.0623e-04, 1.1246e-04], device='cuda:0') +2023-02-07 06:49:30,924 INFO [optim.py:369] (0/4) 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 06:49:38,398 INFO [train.py:901] (0/4) Epoch 24, batch 2500, loss[loss=0.1821, simple_loss=0.2666, pruned_loss=0.04877, over 7977.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2841, pruned_loss=0.05882, over 1614259.60 frames. ], batch size: 21, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:50:02,885 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.19 vs. limit=5.0 +2023-02-07 06:50:11,879 INFO [train.py:901] (0/4) Epoch 24, batch 2550, loss[loss=0.222, simple_loss=0.2891, pruned_loss=0.07745, over 7430.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2837, pruned_loss=0.05896, over 1615505.06 frames. ], batch size: 17, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:50:17,708 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-02-07 06:50:34,558 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3883, 1.6719, 1.6820, 1.0695, 1.6816, 1.4399, 0.2946, 1.6280], + device='cuda:0'), covar=tensor([0.0502, 0.0401, 0.0354, 0.0536, 0.0455, 0.0998, 0.0911, 0.0288], + device='cuda:0'), in_proj_covar=tensor([0.0462, 0.0402, 0.0357, 0.0454, 0.0387, 0.0543, 0.0398, 0.0430], + device='cuda:0'), out_proj_covar=tensor([1.2320e-04, 1.0520e-04, 9.3852e-05, 1.1930e-04, 1.0164e-04, 1.5275e-04, + 1.0707e-04, 1.1345e-04], device='cuda:0') +2023-02-07 06:50:40,457 INFO [optim.py:369] (0/4) 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 06:50:40,654 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.7795, 1.6441, 1.9280, 1.6169, 1.1593, 1.6892, 2.1552, 2.1286], + device='cuda:0'), covar=tensor([0.0496, 0.1276, 0.1675, 0.1457, 0.0602, 0.1467, 0.0669, 0.0600], + device='cuda:0'), in_proj_covar=tensor([0.0098, 0.0152, 0.0189, 0.0159, 0.0100, 0.0163, 0.0112, 0.0144], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:0') +2023-02-07 06:50:41,289 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=188499.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:50:47,911 INFO [train.py:901] (0/4) Epoch 24, batch 2600, loss[loss=0.1562, simple_loss=0.2419, pruned_loss=0.03526, over 7531.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2835, pruned_loss=0.05816, over 1619327.06 frames. ], batch size: 18, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:51:00,092 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.52 vs. limit=2.0 +2023-02-07 06:51:21,917 INFO [train.py:901] (0/4) Epoch 24, batch 2650, loss[loss=0.1815, simple_loss=0.2578, pruned_loss=0.0526, over 7528.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2841, pruned_loss=0.05831, over 1621490.26 frames. ], batch size: 18, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:51:48,597 INFO [optim.py:369] (0/4) 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 06:51:55,401 INFO [train.py:901] (0/4) Epoch 24, batch 2700, loss[loss=0.1619, simple_loss=0.2484, pruned_loss=0.03766, over 7224.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2842, pruned_loss=0.05865, over 1621060.49 frames. ], batch size: 16, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:52:21,721 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 24, batch 2750, loss[loss=0.1924, simple_loss=0.2811, pruned_loss=0.05188, over 8533.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2851, pruned_loss=0.05919, over 1627090.40 frames. ], batch size: 28, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:52:37,532 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 +2023-02-07 06:52:41,607 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-02-07 06:52:44,035 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.1526, 1.3559, 1.5993, 1.3376, 0.7878, 1.4061, 1.2737, 1.0020], + device='cuda:0'), covar=tensor([0.0617, 0.1264, 0.1617, 0.1394, 0.0552, 0.1406, 0.0688, 0.0719], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0153, 0.0189, 0.0160, 0.0100, 0.0162, 0.0112, 0.0144], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:0') +2023-02-07 06:52:57,784 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 24, batch 2800, loss[loss=0.1776, simple_loss=0.2677, pruned_loss=0.04375, over 7923.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2851, pruned_loss=0.05913, over 1621102.72 frames. ], batch size: 20, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:53:40,611 INFO [train.py:901] (0/4) Epoch 24, batch 2850, loss[loss=0.2152, simple_loss=0.2959, pruned_loss=0.06726, over 8180.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.285, pruned_loss=0.05914, over 1620386.76 frames. ], batch size: 23, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:53:42,158 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=188760.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:54:07,924 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 24, batch 2900, loss[loss=0.2434, simple_loss=0.3287, pruned_loss=0.07905, over 8242.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2855, pruned_loss=0.05982, over 1616071.96 frames. ], batch size: 24, lr: 3.16e-03, grad_scale: 8.0 +2023-02-07 06:54:21,524 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 +2023-02-07 06:54:39,464 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=188843.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:54:51,396 INFO [train.py:901] (0/4) Epoch 24, batch 2950, loss[loss=0.1977, simple_loss=0.2901, pruned_loss=0.05263, over 8196.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2846, pruned_loss=0.05922, over 1614020.60 frames. ], batch size: 23, lr: 3.15e-03, grad_scale: 8.0 +2023-02-07 06:54:51,439 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-07 06:55:04,756 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1604, 1.9181, 2.5369, 2.1378, 2.5313, 2.2155, 2.0046, 1.3923], + device='cuda:0'), covar=tensor([0.5545, 0.5068, 0.2023, 0.3720, 0.2560, 0.3182, 0.1958, 0.5588], + device='cuda:0'), in_proj_covar=tensor([0.0949, 0.0993, 0.0816, 0.0958, 0.0997, 0.0907, 0.0757, 0.0833], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 06:55:15,785 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1667, 1.4800, 4.3559, 1.7413, 2.4733, 4.9470, 5.1342, 4.3488], + device='cuda:0'), covar=tensor([0.1293, 0.1963, 0.0289, 0.2249, 0.1187, 0.0212, 0.0585, 0.0564], + device='cuda:0'), in_proj_covar=tensor([0.0299, 0.0323, 0.0285, 0.0316, 0.0315, 0.0270, 0.0426, 0.0303], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 06:55:19,116 INFO [optim.py:369] (0/4) 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 06:55:25,891 INFO [train.py:901] (0/4) Epoch 24, batch 3000, loss[loss=0.1854, simple_loss=0.2796, pruned_loss=0.04561, over 7796.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2856, pruned_loss=0.05934, over 1616601.35 frames. ], batch size: 20, lr: 3.15e-03, grad_scale: 8.0 +2023-02-07 06:55:25,891 INFO [train.py:926] (0/4) Computing validation loss +2023-02-07 06:55:39,548 INFO [train.py:935] (0/4) Epoch 24, validation: loss=0.1724, simple_loss=0.2726, pruned_loss=0.03604, over 944034.00 frames. +2023-02-07 06:55:39,549 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-07 06:55:46,717 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 +2023-02-07 06:56:05,990 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=188947.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:56:07,356 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=188949.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:56:13,960 INFO [train.py:901] (0/4) Epoch 24, batch 3050, loss[loss=0.2076, simple_loss=0.2993, pruned_loss=0.05798, over 8366.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2858, pruned_loss=0.05979, over 1617013.34 frames. ], batch size: 24, lr: 3.15e-03, grad_scale: 8.0 +2023-02-07 06:56:14,161 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=188958.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:56:41,562 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 24, batch 3100, loss[loss=0.2181, simple_loss=0.2983, pruned_loss=0.06899, over 8029.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2868, pruned_loss=0.06072, over 1614523.83 frames. ], batch size: 22, lr: 3.15e-03, grad_scale: 8.0 +2023-02-07 06:56:54,797 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189016.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:57:12,097 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189041.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:57:16,965 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189048.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:57:23,538 INFO [train.py:901] (0/4) Epoch 24, batch 3150, loss[loss=0.2136, simple_loss=0.2974, pruned_loss=0.06485, over 7983.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2875, pruned_loss=0.0611, over 1615274.51 frames. ], batch size: 21, lr: 3.15e-03, grad_scale: 8.0 +2023-02-07 06:57:40,667 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189082.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:57:50,859 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189097.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:57:51,169 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 +2023-02-07 06:57:51,369 INFO [optim.py:369] (0/4) 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 06:57:59,743 INFO [train.py:901] (0/4) Epoch 24, batch 3200, loss[loss=0.1746, simple_loss=0.2612, pruned_loss=0.04396, over 8356.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2869, pruned_loss=0.06092, over 1612945.51 frames. ], batch size: 24, lr: 3.15e-03, grad_scale: 8.0 +2023-02-07 06:58:33,957 INFO [train.py:901] (0/4) Epoch 24, batch 3250, loss[loss=0.2324, simple_loss=0.3081, pruned_loss=0.0784, over 8497.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.287, pruned_loss=0.06094, over 1614968.66 frames. ], batch size: 28, lr: 3.15e-03, grad_scale: 16.0 +2023-02-07 06:59:01,531 INFO [optim.py:369] (0/4) 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 06:59:08,533 INFO [train.py:901] (0/4) Epoch 24, batch 3300, loss[loss=0.1806, simple_loss=0.275, pruned_loss=0.04309, over 8608.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2858, pruned_loss=0.06047, over 1615745.90 frames. ], batch size: 34, lr: 3.15e-03, grad_scale: 16.0 +2023-02-07 06:59:10,274 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 +2023-02-07 06:59:12,860 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189214.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:59:13,785 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 +2023-02-07 06:59:26,873 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8485, 1.4329, 1.6326, 1.3185, 0.9809, 1.4936, 1.6701, 1.4884], + device='cuda:0'), covar=tensor([0.0516, 0.1248, 0.1635, 0.1448, 0.0603, 0.1420, 0.0664, 0.0630], + device='cuda:0'), in_proj_covar=tensor([0.0098, 0.0154, 0.0190, 0.0160, 0.0101, 0.0163, 0.0112, 0.0145], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:0') +2023-02-07 06:59:30,912 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189239.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:59:37,688 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1489, 2.0092, 2.5826, 2.1381, 2.6211, 2.2218, 2.0046, 1.4450], + device='cuda:0'), covar=tensor([0.5699, 0.5029, 0.1995, 0.3817, 0.2441, 0.3155, 0.1979, 0.5398], + device='cuda:0'), in_proj_covar=tensor([0.0941, 0.0987, 0.0808, 0.0951, 0.0991, 0.0900, 0.0751, 0.0825], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 06:59:44,169 INFO [train.py:901] (0/4) Epoch 24, batch 3350, loss[loss=0.192, simple_loss=0.272, pruned_loss=0.05602, over 7723.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2866, pruned_loss=0.06048, over 1622266.01 frames. ], batch size: 18, lr: 3.15e-03, grad_scale: 16.0 +2023-02-07 06:59:49,403 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189266.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 06:59:56,566 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4759, 1.3819, 1.8274, 1.2313, 1.1320, 1.7422, 0.1814, 1.1753], + device='cuda:0'), covar=tensor([0.1569, 0.1398, 0.0428, 0.1076, 0.2588, 0.0566, 0.2172, 0.1321], + device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0200, 0.0131, 0.0221, 0.0271, 0.0138, 0.0171, 0.0195], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 07:00:05,776 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=189291.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:00:07,043 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=189293.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:00:10,342 INFO [optim.py:369] (0/4) 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 07:00:17,748 INFO [train.py:901] (0/4) Epoch 24, batch 3400, loss[loss=0.1969, simple_loss=0.2823, pruned_loss=0.05573, over 8466.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2873, pruned_loss=0.06122, over 1621052.20 frames. ], batch size: 25, lr: 3.15e-03, grad_scale: 16.0 +2023-02-07 07:00:52,454 INFO [train.py:901] (0/4) Epoch 24, batch 3450, loss[loss=0.1995, simple_loss=0.2804, pruned_loss=0.05927, over 8188.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2871, pruned_loss=0.06083, over 1624250.16 frames. ], batch size: 23, lr: 3.15e-03, grad_scale: 16.0 +2023-02-07 07:00:57,502 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189365.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:01:16,585 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=189392.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:01:20,612 INFO [optim.py:369] (0/4) 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 07:01:26,250 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189406.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:01:26,979 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.5996, 2.3726, 3.0808, 2.5544, 3.0477, 2.6192, 2.4820, 1.9045], + device='cuda:0'), covar=tensor([0.5551, 0.4944, 0.2043, 0.3996, 0.2614, 0.3017, 0.1803, 0.5563], + device='cuda:0'), in_proj_covar=tensor([0.0940, 0.0988, 0.0808, 0.0954, 0.0993, 0.0901, 0.0752, 0.0825], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 07:01:27,411 INFO [train.py:901] (0/4) Epoch 24, batch 3500, loss[loss=0.1687, simple_loss=0.2479, pruned_loss=0.04478, over 7510.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2869, pruned_loss=0.06115, over 1621006.99 frames. ], batch size: 18, lr: 3.15e-03, grad_scale: 16.0 +2023-02-07 07:01:27,606 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189408.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:01:40,527 WARNING [train.py:1067] (0/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-07 07:01:40,594 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=189426.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:01:50,792 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=189441.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:01:58,888 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.5653, 1.6641, 4.7678, 1.9841, 4.3197, 3.9990, 4.3277, 4.1752], + device='cuda:0'), covar=tensor([0.0566, 0.4362, 0.0479, 0.4014, 0.1006, 0.0892, 0.0523, 0.0683], + device='cuda:0'), in_proj_covar=tensor([0.0647, 0.0654, 0.0714, 0.0643, 0.0722, 0.0619, 0.0615, 0.0694], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 07:02:03,504 INFO [train.py:901] (0/4) Epoch 24, batch 3550, loss[loss=0.2087, simple_loss=0.2937, pruned_loss=0.06184, over 8435.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2875, pruned_loss=0.06108, over 1622596.34 frames. ], batch size: 48, lr: 3.15e-03, grad_scale: 16.0 +2023-02-07 07:02:31,306 INFO [optim.py:369] (0/4) 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 07:02:37,626 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189507.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:02:38,143 INFO [train.py:901] (0/4) Epoch 24, batch 3600, loss[loss=0.1815, simple_loss=0.2763, pruned_loss=0.04333, over 8460.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2862, pruned_loss=0.06082, over 1612815.95 frames. ], batch size: 27, lr: 3.15e-03, grad_scale: 16.0 +2023-02-07 07:02:45,840 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189519.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:02:47,253 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.5507, 1.6330, 4.7702, 1.7401, 4.2555, 3.9489, 4.3360, 4.2273], + device='cuda:0'), covar=tensor([0.0663, 0.4564, 0.0482, 0.4274, 0.1049, 0.0935, 0.0579, 0.0664], + device='cuda:0'), in_proj_covar=tensor([0.0648, 0.0656, 0.0714, 0.0645, 0.0724, 0.0621, 0.0617, 0.0695], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 07:03:01,844 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189541.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:03:12,181 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189556.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:03:13,420 INFO [train.py:901] (0/4) Epoch 24, batch 3650, loss[loss=0.2103, simple_loss=0.304, pruned_loss=0.05829, over 8490.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.287, pruned_loss=0.06158, over 1613293.25 frames. ], batch size: 26, lr: 3.15e-03, grad_scale: 16.0 +2023-02-07 07:03:14,316 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6972, 2.0164, 2.0934, 1.3903, 2.2046, 1.4408, 0.7733, 1.9487], + device='cuda:0'), covar=tensor([0.0688, 0.0384, 0.0313, 0.0637, 0.0437, 0.0955, 0.0941, 0.0365], + device='cuda:0'), in_proj_covar=tensor([0.0458, 0.0398, 0.0354, 0.0450, 0.0385, 0.0539, 0.0394, 0.0427], + device='cuda:0'), out_proj_covar=tensor([1.2221e-04, 1.0402e-04, 9.2805e-05, 1.1810e-04, 1.0125e-04, 1.5143e-04, + 1.0595e-04, 1.1258e-04], device='cuda:0') +2023-02-07 07:03:41,145 WARNING [train.py:1067] (0/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-07 07:03:41,748 INFO [optim.py:369] (0/4) 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 07:03:48,375 INFO [train.py:901] (0/4) Epoch 24, batch 3700, loss[loss=0.1858, simple_loss=0.2626, pruned_loss=0.05449, over 7202.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2866, pruned_loss=0.06149, over 1608462.50 frames. ], batch size: 16, lr: 3.15e-03, grad_scale: 16.0 +2023-02-07 07:03:49,804 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=189610.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:03:56,018 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0995, 1.8009, 2.2939, 1.9272, 2.2885, 2.1376, 1.9524, 1.1520], + device='cuda:0'), covar=tensor([0.5675, 0.5194, 0.2093, 0.3760, 0.2585, 0.3279, 0.1947, 0.5305], + device='cuda:0'), in_proj_covar=tensor([0.0945, 0.0993, 0.0811, 0.0960, 0.0999, 0.0906, 0.0755, 0.0830], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 07:04:10,923 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5799, 1.2773, 1.6478, 1.2515, 0.9514, 1.4471, 1.4796, 1.3090], + device='cuda:0'), covar=tensor([0.0574, 0.1271, 0.1641, 0.1532, 0.0580, 0.1454, 0.0714, 0.0673], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0153, 0.0189, 0.0159, 0.0100, 0.0162, 0.0112, 0.0144], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:0') +2023-02-07 07:04:23,135 INFO [train.py:901] (0/4) Epoch 24, batch 3750, loss[loss=0.1906, simple_loss=0.282, pruned_loss=0.04959, over 8295.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2867, pruned_loss=0.06118, over 1610940.48 frames. ], batch size: 23, lr: 3.15e-03, grad_scale: 16.0 +2023-02-07 07:04:23,273 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189658.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:04:26,007 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189662.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:04:27,231 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189664.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:04:42,767 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189687.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:04:43,971 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189689.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:04:51,128 INFO [optim.py:369] (0/4) 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 07:04:51,254 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.5578, 4.4863, 4.0182, 2.2583, 3.9409, 4.1862, 4.0160, 3.9687], + device='cuda:0'), covar=tensor([0.0692, 0.0513, 0.1023, 0.4377, 0.0864, 0.0883, 0.1233, 0.0774], + device='cuda:0'), in_proj_covar=tensor([0.0523, 0.0441, 0.0431, 0.0537, 0.0427, 0.0444, 0.0423, 0.0387], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 07:04:57,807 INFO [train.py:901] (0/4) Epoch 24, batch 3800, loss[loss=0.2404, simple_loss=0.3151, pruned_loss=0.08284, over 8442.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2863, pruned_loss=0.06064, over 1612025.25 frames. ], batch size: 27, lr: 3.15e-03, grad_scale: 16.0 +2023-02-07 07:04:58,579 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=189709.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:05:07,483 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189722.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:05:09,540 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189725.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:05:24,350 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189746.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:05:27,567 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6288, 1.9349, 2.9310, 1.4757, 2.1740, 2.1354, 1.6789, 2.1820], + device='cuda:0'), covar=tensor([0.2090, 0.2864, 0.0922, 0.4983, 0.2085, 0.3321, 0.2602, 0.2416], + device='cuda:0'), in_proj_covar=tensor([0.0530, 0.0618, 0.0559, 0.0657, 0.0655, 0.0602, 0.0550, 0.0640], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 07:05:32,056 INFO [train.py:901] (0/4) Epoch 24, batch 3850, loss[loss=0.1812, simple_loss=0.2692, pruned_loss=0.04661, over 8240.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2853, pruned_loss=0.05978, over 1611395.62 frames. ], batch size: 22, lr: 3.15e-03, grad_scale: 16.0 +2023-02-07 07:05:35,651 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189763.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:05:47,651 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-07 07:05:53,006 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189788.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:05:59,077 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189797.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:05:59,532 INFO [optim.py:369] (0/4) 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 07:06:02,174 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 +2023-02-07 07:06:06,335 INFO [train.py:901] (0/4) Epoch 24, batch 3900, loss[loss=0.1798, simple_loss=0.2765, pruned_loss=0.04157, over 8509.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2864, pruned_loss=0.06031, over 1612032.60 frames. ], batch size: 28, lr: 3.15e-03, grad_scale: 16.0 +2023-02-07 07:06:10,103 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189812.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:06:17,538 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189822.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:06:18,887 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189824.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:06:27,531 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189837.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:06:42,140 INFO [train.py:901] (0/4) Epoch 24, batch 3950, loss[loss=0.2059, simple_loss=0.2743, pruned_loss=0.06871, over 7922.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2843, pruned_loss=0.05965, over 1608193.98 frames. ], batch size: 20, lr: 3.15e-03, grad_scale: 16.0 +2023-02-07 07:06:45,540 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=189863.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:07:09,585 INFO [optim.py:369] (0/4) 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 07:07:14,574 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.21 vs. limit=5.0 +2023-02-07 07:07:16,315 INFO [train.py:901] (0/4) Epoch 24, batch 4000, loss[loss=0.2075, simple_loss=0.2958, pruned_loss=0.05958, over 8196.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2848, pruned_loss=0.05983, over 1610080.74 frames. ], batch size: 23, lr: 3.15e-03, grad_scale: 16.0 +2023-02-07 07:07:45,303 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.3809, 1.6479, 4.5432, 1.6719, 4.0572, 3.7475, 4.0838, 3.9962], + device='cuda:0'), covar=tensor([0.0634, 0.4581, 0.0528, 0.4359, 0.1070, 0.0978, 0.0661, 0.0666], + device='cuda:0'), in_proj_covar=tensor([0.0652, 0.0660, 0.0717, 0.0646, 0.0724, 0.0624, 0.0622, 0.0696], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 07:07:51,144 INFO [train.py:901] (0/4) Epoch 24, batch 4050, loss[loss=0.2464, simple_loss=0.3338, pruned_loss=0.07951, over 8564.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2845, pruned_loss=0.05981, over 1612124.33 frames. ], batch size: 31, lr: 3.15e-03, grad_scale: 16.0 +2023-02-07 07:08:05,463 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189978.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:08:07,495 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189981.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:08:18,677 INFO [optim.py:369] (0/4) 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 07:08:18,872 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.1553, 1.2628, 1.5471, 1.2128, 0.7607, 1.3527, 1.1793, 0.9027], + device='cuda:0'), covar=tensor([0.0653, 0.1321, 0.1621, 0.1530, 0.0600, 0.1536, 0.0752, 0.0771], + device='cuda:0'), in_proj_covar=tensor([0.0098, 0.0153, 0.0189, 0.0160, 0.0100, 0.0163, 0.0112, 0.0144], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:0') +2023-02-07 07:08:20,165 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-190000.pt +2023-02-07 07:08:22,559 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=190002.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:08:25,394 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190006.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:08:27,378 INFO [train.py:901] (0/4) Epoch 24, batch 4100, loss[loss=0.2128, simple_loss=0.2889, pruned_loss=0.0684, over 8344.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.284, pruned_loss=0.05942, over 1610336.08 frames. ], batch size: 26, lr: 3.15e-03, grad_scale: 16.0 +2023-02-07 07:09:02,436 INFO [train.py:901] (0/4) Epoch 24, batch 4150, loss[loss=0.2725, simple_loss=0.3463, pruned_loss=0.09941, over 8627.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2853, pruned_loss=0.06001, over 1609909.74 frames. ], batch size: 34, lr: 3.15e-03, grad_scale: 16.0 +2023-02-07 07:09:08,092 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=190066.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:09:13,709 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8074, 1.4466, 3.2488, 1.4146, 2.2380, 3.5240, 3.7178, 3.0106], + device='cuda:0'), covar=tensor([0.1295, 0.1824, 0.0351, 0.2204, 0.1027, 0.0251, 0.0556, 0.0566], + device='cuda:0'), in_proj_covar=tensor([0.0298, 0.0323, 0.0286, 0.0315, 0.0314, 0.0271, 0.0427, 0.0303], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 07:09:17,887 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190080.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:09:25,157 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 07:09:35,501 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190105.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:09:37,391 INFO [train.py:901] (0/4) Epoch 24, batch 4200, loss[loss=0.2279, simple_loss=0.3093, pruned_loss=0.07319, over 8722.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2854, pruned_loss=0.06013, over 1609558.74 frames. ], batch size: 49, lr: 3.14e-03, grad_scale: 16.0 +2023-02-07 07:09:43,677 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190117.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:09:48,186 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-07 07:10:10,655 WARNING [train.py:1067] (0/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-07 07:10:11,421 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190157.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 07:10:11,874 INFO [train.py:901] (0/4) Epoch 24, batch 4250, loss[loss=0.2562, simple_loss=0.3339, pruned_loss=0.08925, over 8562.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2863, pruned_loss=0.06072, over 1612059.17 frames. ], batch size: 34, lr: 3.14e-03, grad_scale: 16.0 +2023-02-07 07:10:19,337 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190169.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:10:28,694 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190181.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:10:40,140 INFO [optim.py:369] (0/4) 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 07:10:45,750 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190205.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:10:47,629 INFO [train.py:901] (0/4) Epoch 24, batch 4300, loss[loss=0.2142, simple_loss=0.2994, pruned_loss=0.0645, over 8482.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2872, pruned_loss=0.06035, over 1620213.43 frames. ], batch size: 29, lr: 3.14e-03, grad_scale: 16.0 +2023-02-07 07:10:56,873 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 +2023-02-07 07:11:05,306 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 24, batch 4350, loss[loss=0.1867, simple_loss=0.282, pruned_loss=0.04566, over 8110.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2866, pruned_loss=0.06018, over 1618356.67 frames. ], batch size: 23, lr: 3.14e-03, grad_scale: 16.0 +2023-02-07 07:11:22,126 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8404, 1.7177, 2.3666, 1.6247, 1.4127, 2.3135, 0.3746, 1.4366], + device='cuda:0'), covar=tensor([0.1499, 0.1324, 0.0345, 0.0976, 0.2294, 0.0424, 0.1949, 0.1202], + device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0201, 0.0130, 0.0222, 0.0271, 0.0138, 0.0171, 0.0195], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 07:11:22,812 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190259.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:11:40,181 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-07 07:11:48,230 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 +2023-02-07 07:11:50,408 INFO [optim.py:369] (0/4) 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 07:11:57,331 INFO [train.py:901] (0/4) Epoch 24, batch 4400, loss[loss=0.2404, simple_loss=0.3132, pruned_loss=0.08381, over 8361.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.286, pruned_loss=0.05988, over 1619209.26 frames. ], batch size: 24, lr: 3.14e-03, grad_scale: 16.0 +2023-02-07 07:12:15,560 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190334.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:12:23,152 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-07 07:12:32,980 INFO [train.py:901] (0/4) Epoch 24, batch 4450, loss[loss=0.211, simple_loss=0.3005, pruned_loss=0.0607, over 8338.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2858, pruned_loss=0.05937, over 1618836.90 frames. ], batch size: 26, lr: 3.14e-03, grad_scale: 16.0 +2023-02-07 07:12:43,417 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 07:13:00,471 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190398.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:13:07,778 INFO [train.py:901] (0/4) Epoch 24, batch 4500, loss[loss=0.2058, simple_loss=0.289, pruned_loss=0.06134, over 8636.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2859, pruned_loss=0.05964, over 1620972.05 frames. ], batch size: 34, lr: 3.14e-03, grad_scale: 16.0 +2023-02-07 07:13:14,918 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190417.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:13:17,415 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-07 07:13:29,010 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190437.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:13:29,617 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2011, 1.3515, 1.5953, 1.2844, 0.7325, 1.4432, 1.2131, 1.0733], + device='cuda:0'), covar=tensor([0.0596, 0.1253, 0.1627, 0.1436, 0.0551, 0.1423, 0.0668, 0.0709], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0153, 0.0189, 0.0160, 0.0100, 0.0162, 0.0112, 0.0144], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:0') +2023-02-07 07:13:42,697 INFO [train.py:901] (0/4) Epoch 24, batch 4550, loss[loss=0.2109, simple_loss=0.3035, pruned_loss=0.05917, over 8245.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.286, pruned_loss=0.06026, over 1618652.09 frames. ], batch size: 24, lr: 3.14e-03, grad_scale: 16.0 +2023-02-07 07:13:44,857 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190461.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:13:45,491 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190462.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:13:46,754 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5239, 2.4713, 1.7882, 2.4228, 2.1621, 1.4508, 2.1292, 2.2244], + device='cuda:0'), covar=tensor([0.1561, 0.0472, 0.1349, 0.0556, 0.0823, 0.1754, 0.1025, 0.0976], + device='cuda:0'), in_proj_covar=tensor([0.0355, 0.0234, 0.0337, 0.0311, 0.0300, 0.0343, 0.0346, 0.0318], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 07:13:54,378 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8444, 1.5690, 3.1570, 1.5121, 2.3217, 3.3847, 3.5196, 2.9051], + device='cuda:0'), covar=tensor([0.1181, 0.1694, 0.0323, 0.2078, 0.0940, 0.0255, 0.0619, 0.0564], + device='cuda:0'), in_proj_covar=tensor([0.0298, 0.0323, 0.0285, 0.0316, 0.0314, 0.0271, 0.0428, 0.0303], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 07:14:02,589 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190486.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:14:10,717 INFO [optim.py:369] (0/4) 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 07:14:12,830 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=190501.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 07:14:17,566 INFO [train.py:901] (0/4) Epoch 24, batch 4600, loss[loss=0.2335, simple_loss=0.3121, pruned_loss=0.07746, over 8591.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2866, pruned_loss=0.06056, over 1617621.45 frames. ], batch size: 34, lr: 3.14e-03, grad_scale: 16.0 +2023-02-07 07:14:21,247 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=190513.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:14:27,617 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0411, 1.7406, 3.2142, 1.6029, 2.3744, 3.5236, 3.6497, 3.0197], + device='cuda:0'), covar=tensor([0.1169, 0.1656, 0.0386, 0.2114, 0.1066, 0.0279, 0.0635, 0.0614], + device='cuda:0'), in_proj_covar=tensor([0.0297, 0.0323, 0.0285, 0.0315, 0.0314, 0.0270, 0.0428, 0.0303], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 07:14:35,973 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6493, 2.2564, 4.0708, 1.4028, 2.8025, 2.1379, 1.8567, 2.7275], + device='cuda:0'), covar=tensor([0.2093, 0.2697, 0.0832, 0.4960, 0.2074, 0.3399, 0.2487, 0.2828], + device='cuda:0'), in_proj_covar=tensor([0.0535, 0.0622, 0.0562, 0.0661, 0.0658, 0.0605, 0.0553, 0.0644], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 07:14:54,265 INFO [train.py:901] (0/4) Epoch 24, batch 4650, loss[loss=0.1843, simple_loss=0.2643, pruned_loss=0.05212, over 7809.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2856, pruned_loss=0.05994, over 1617002.62 frames. ], batch size: 20, lr: 3.14e-03, grad_scale: 16.0 +2023-02-07 07:15:19,160 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8056, 1.5711, 1.6773, 1.5286, 1.0992, 1.6084, 1.7135, 1.5500], + device='cuda:0'), covar=tensor([0.0568, 0.0955, 0.1273, 0.1154, 0.0578, 0.1123, 0.0703, 0.0521], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0152, 0.0188, 0.0160, 0.0100, 0.0162, 0.0111, 0.0144], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:0') +2023-02-07 07:15:22,378 INFO [optim.py:369] (0/4) 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 07:15:29,192 INFO [train.py:901] (0/4) Epoch 24, batch 4700, loss[loss=0.2357, simple_loss=0.3192, pruned_loss=0.07609, over 8360.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2862, pruned_loss=0.06024, over 1616757.08 frames. ], batch size: 24, lr: 3.14e-03, grad_scale: 16.0 +2023-02-07 07:15:29,951 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190609.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 07:15:34,413 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190616.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 07:15:41,741 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190627.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:15:42,469 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190628.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:15:57,393 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8691, 1.5955, 3.3493, 1.5784, 2.3198, 3.6394, 3.7542, 3.1149], + device='cuda:0'), covar=tensor([0.1186, 0.1667, 0.0303, 0.1991, 0.0935, 0.0238, 0.0615, 0.0553], + device='cuda:0'), in_proj_covar=tensor([0.0295, 0.0322, 0.0283, 0.0314, 0.0313, 0.0270, 0.0426, 0.0301], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 07:16:04,013 INFO [train.py:901] (0/4) Epoch 24, batch 4750, loss[loss=0.2183, simple_loss=0.2878, pruned_loss=0.07442, over 7525.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2855, pruned_loss=0.06021, over 1613861.56 frames. ], batch size: 18, lr: 3.14e-03, grad_scale: 16.0 +2023-02-07 07:16:18,692 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=190678.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:16:20,754 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-07 07:16:22,902 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-07 07:16:29,407 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190693.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:16:32,683 INFO [optim.py:369] (0/4) 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 07:16:40,394 INFO [train.py:901] (0/4) Epoch 24, batch 4800, loss[loss=0.2317, simple_loss=0.3104, pruned_loss=0.07652, over 6672.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.285, pruned_loss=0.05986, over 1610559.89 frames. ], batch size: 71, lr: 3.14e-03, grad_scale: 16.0 +2023-02-07 07:17:13,012 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-07 07:17:15,041 INFO [train.py:901] (0/4) Epoch 24, batch 4850, loss[loss=0.1862, simple_loss=0.2755, pruned_loss=0.04848, over 8667.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.285, pruned_loss=0.05981, over 1613027.48 frames. ], batch size: 34, lr: 3.14e-03, grad_scale: 16.0 +2023-02-07 07:17:15,277 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6869, 2.1247, 3.1989, 1.4952, 2.4842, 2.0589, 1.8734, 2.4511], + device='cuda:0'), covar=tensor([0.2019, 0.2795, 0.0868, 0.4890, 0.2012, 0.3514, 0.2550, 0.2344], + device='cuda:0'), in_proj_covar=tensor([0.0525, 0.0612, 0.0552, 0.0650, 0.0646, 0.0595, 0.0545, 0.0632], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 07:17:17,934 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=190761.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:17:40,153 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190793.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:17:43,508 INFO [optim.py:369] (0/4) 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 07:17:50,723 INFO [train.py:901] (0/4) Epoch 24, batch 4900, loss[loss=0.1834, simple_loss=0.2561, pruned_loss=0.05538, over 7534.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2838, pruned_loss=0.0589, over 1615267.20 frames. ], batch size: 18, lr: 3.14e-03, grad_scale: 16.0 +2023-02-07 07:18:09,671 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.5610, 1.6034, 4.7784, 1.8492, 4.2342, 4.0166, 4.3173, 4.2026], + device='cuda:0'), covar=tensor([0.0569, 0.4399, 0.0494, 0.4000, 0.1042, 0.0856, 0.0527, 0.0640], + device='cuda:0'), in_proj_covar=tensor([0.0646, 0.0653, 0.0713, 0.0641, 0.0721, 0.0619, 0.0618, 0.0692], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 07:18:25,854 INFO [train.py:901] (0/4) Epoch 24, batch 4950, loss[loss=0.1499, simple_loss=0.2397, pruned_loss=0.03001, over 6391.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2837, pruned_loss=0.05887, over 1615458.17 frames. ], batch size: 14, lr: 3.14e-03, grad_scale: 16.0 +2023-02-07 07:18:36,138 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190884.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:18:53,847 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190897.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 07:18:54,306 INFO [optim.py:369] (0/4) 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 07:19:01,826 INFO [train.py:901] (0/4) Epoch 24, batch 5000, loss[loss=0.2108, simple_loss=0.3042, pruned_loss=0.05868, over 8458.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.284, pruned_loss=0.05853, over 1616667.06 frames. ], batch size: 25, lr: 3.14e-03, grad_scale: 16.0 +2023-02-07 07:19:02,718 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190909.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:19:33,474 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=190953.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 07:19:36,598 INFO [train.py:901] (0/4) Epoch 24, batch 5050, loss[loss=0.205, simple_loss=0.2909, pruned_loss=0.05957, over 8514.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2831, pruned_loss=0.05786, over 1610449.62 frames. ], batch size: 28, lr: 3.14e-03, grad_scale: 16.0 +2023-02-07 07:19:45,288 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=190971.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:19:51,188 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-07 07:20:04,905 INFO [optim.py:369] (0/4) 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 07:20:11,615 INFO [train.py:901] (0/4) Epoch 24, batch 5100, loss[loss=0.207, simple_loss=0.2967, pruned_loss=0.05867, over 8437.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2834, pruned_loss=0.05757, over 1614565.38 frames. ], batch size: 29, lr: 3.14e-03, grad_scale: 16.0 +2023-02-07 07:20:25,285 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9680, 1.7132, 3.3488, 1.7369, 2.5194, 3.6912, 3.7321, 3.1849], + device='cuda:0'), covar=tensor([0.1188, 0.1693, 0.0337, 0.1958, 0.1049, 0.0231, 0.0603, 0.0522], + device='cuda:0'), in_proj_covar=tensor([0.0296, 0.0322, 0.0284, 0.0314, 0.0313, 0.0271, 0.0428, 0.0301], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 07:20:31,893 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=191037.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:20:40,071 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=191049.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:20:44,519 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 +2023-02-07 07:20:46,726 INFO [train.py:901] (0/4) Epoch 24, batch 5150, loss[loss=0.233, simple_loss=0.3152, pruned_loss=0.07543, over 8312.00 frames. ], tot_loss[loss=0.199, simple_loss=0.283, pruned_loss=0.05753, over 1616619.25 frames. ], batch size: 25, lr: 3.14e-03, grad_scale: 16.0 +2023-02-07 07:20:53,679 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191068.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 07:20:57,806 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=191074.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:21:03,775 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3156, 2.1020, 1.7156, 1.9792, 1.8716, 1.4654, 1.7064, 1.6586], + device='cuda:0'), covar=tensor([0.1216, 0.0436, 0.1152, 0.0494, 0.0658, 0.1549, 0.0935, 0.0809], + device='cuda:0'), in_proj_covar=tensor([0.0356, 0.0234, 0.0337, 0.0309, 0.0299, 0.0341, 0.0346, 0.0318], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 07:21:05,748 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=191086.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:21:05,783 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) Epoch 24, batch 5200, loss[loss=0.1879, simple_loss=0.2797, pruned_loss=0.04807, over 8198.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2827, pruned_loss=0.05834, over 1611851.39 frames. ], batch size: 23, lr: 3.14e-03, grad_scale: 16.0 +2023-02-07 07:21:38,118 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=191132.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:21:45,746 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.61 vs. limit=2.0 +2023-02-07 07:21:50,039 WARNING [train.py:1067] (0/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-07 07:21:50,828 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=191150.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:21:52,265 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191152.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:21:55,620 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=191157.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:21:56,110 INFO [train.py:901] (0/4) Epoch 24, batch 5250, loss[loss=0.2553, simple_loss=0.3311, pruned_loss=0.08972, over 8329.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2842, pruned_loss=0.05901, over 1611959.82 frames. ], batch size: 26, lr: 3.14e-03, grad_scale: 8.0 +2023-02-07 07:22:25,868 INFO [optim.py:369] (0/4) 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 07:22:28,326 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=191203.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:22:31,606 INFO [train.py:901] (0/4) Epoch 24, batch 5300, loss[loss=0.2025, simple_loss=0.2909, pruned_loss=0.05705, over 7809.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2837, pruned_loss=0.05887, over 1602287.62 frames. ], batch size: 20, lr: 3.14e-03, grad_scale: 8.0 +2023-02-07 07:23:07,066 INFO [train.py:901] (0/4) Epoch 24, batch 5350, loss[loss=0.2145, simple_loss=0.297, pruned_loss=0.06594, over 8527.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2847, pruned_loss=0.05931, over 1607466.33 frames. ], batch size: 39, lr: 3.14e-03, grad_scale: 8.0 +2023-02-07 07:23:36,917 INFO [optim.py:369] (0/4) 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 07:23:42,381 INFO [train.py:901] (0/4) Epoch 24, batch 5400, loss[loss=0.2103, simple_loss=0.2951, pruned_loss=0.06277, over 8431.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2836, pruned_loss=0.05909, over 1604282.75 frames. ], batch size: 27, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:23:53,139 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=191324.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 07:24:05,735 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=191342.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:24:11,129 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 24, batch 5450, loss[loss=0.2102, simple_loss=0.2954, pruned_loss=0.0625, over 8601.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2843, pruned_loss=0.05967, over 1608867.59 frames. ], batch size: 31, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:24:20,828 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-07 07:24:23,274 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=191367.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:24:34,960 INFO [zipformer.py:1185] (0/4) 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] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-07 07:24:46,091 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 24, batch 5500, loss[loss=0.1793, simple_loss=0.2539, pruned_loss=0.05234, over 7292.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2842, pruned_loss=0.05993, over 1611917.67 frames. ], batch size: 16, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:24:52,648 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=191408.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:24:59,653 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8612, 1.5262, 3.5393, 1.6442, 2.4513, 3.8285, 3.9293, 3.2518], + device='cuda:0'), covar=tensor([0.1218, 0.1786, 0.0268, 0.1928, 0.0899, 0.0218, 0.0518, 0.0547], + device='cuda:0'), in_proj_covar=tensor([0.0298, 0.0325, 0.0287, 0.0316, 0.0315, 0.0273, 0.0431, 0.0304], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 07:25:07,779 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=191430.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:25:10,085 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=191433.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:25:12,206 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7054, 1.9128, 2.1064, 1.4175, 2.2320, 1.4448, 0.6657, 1.8429], + device='cuda:0'), covar=tensor([0.0699, 0.0413, 0.0301, 0.0651, 0.0486, 0.1026, 0.1010, 0.0387], + device='cuda:0'), in_proj_covar=tensor([0.0463, 0.0404, 0.0359, 0.0456, 0.0388, 0.0542, 0.0401, 0.0434], + device='cuda:0'), out_proj_covar=tensor([1.2355e-04, 1.0545e-04, 9.4180e-05, 1.1973e-04, 1.0200e-04, 1.5191e-04, + 1.0759e-04, 1.1450e-04], device='cuda:0') +2023-02-07 07:25:28,136 INFO [train.py:901] (0/4) Epoch 24, batch 5550, loss[loss=0.2205, simple_loss=0.3086, pruned_loss=0.06619, over 8458.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2854, pruned_loss=0.0601, over 1614863.77 frames. ], batch size: 25, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:25:53,333 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 07:26:00,914 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3415, 2.1004, 2.5699, 2.2257, 2.5357, 2.2848, 2.2011, 1.8684], + device='cuda:0'), covar=tensor([0.4076, 0.4051, 0.1794, 0.3148, 0.2273, 0.2700, 0.1587, 0.4109], + device='cuda:0'), in_proj_covar=tensor([0.0946, 0.0994, 0.0813, 0.0964, 0.1002, 0.0907, 0.0756, 0.0832], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 07:26:03,389 INFO [train.py:901] (0/4) Epoch 24, batch 5600, loss[loss=0.2065, simple_loss=0.287, pruned_loss=0.06304, over 8481.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2862, pruned_loss=0.06108, over 1609905.79 frames. ], batch size: 48, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:26:29,646 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191545.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:26:30,939 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=191547.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:26:38,448 INFO [train.py:901] (0/4) Epoch 24, batch 5650, loss[loss=0.2192, simple_loss=0.2941, pruned_loss=0.07211, over 7652.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2865, pruned_loss=0.06108, over 1608922.50 frames. ], batch size: 19, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:26:45,404 WARNING [train.py:1067] (0/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-07 07:26:49,860 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1475, 1.8439, 2.2754, 2.0063, 2.2876, 2.1552, 1.9662, 1.0756], + device='cuda:0'), covar=tensor([0.5271, 0.5104, 0.2167, 0.3648, 0.2443, 0.3167, 0.1982, 0.5403], + device='cuda:0'), in_proj_covar=tensor([0.0948, 0.0996, 0.0815, 0.0966, 0.1003, 0.0908, 0.0757, 0.0832], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 07:27:08,572 INFO [optim.py:369] (0/4) 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 07:27:13,761 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7846, 2.2502, 3.7740, 1.6592, 2.9413, 2.3278, 1.8668, 2.7747], + device='cuda:0'), covar=tensor([0.1920, 0.2777, 0.0923, 0.4575, 0.1816, 0.3115, 0.2340, 0.2512], + device='cuda:0'), in_proj_covar=tensor([0.0532, 0.0617, 0.0556, 0.0656, 0.0653, 0.0600, 0.0548, 0.0637], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 07:27:14,234 INFO [train.py:901] (0/4) Epoch 24, batch 5700, loss[loss=0.2078, simple_loss=0.2965, pruned_loss=0.05951, over 8711.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2862, pruned_loss=0.06056, over 1614125.22 frames. ], batch size: 39, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:27:15,039 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191609.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:27:37,282 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 +2023-02-07 07:27:49,744 INFO [train.py:901] (0/4) Epoch 24, batch 5750, loss[loss=0.2052, simple_loss=0.2809, pruned_loss=0.06477, over 7644.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2863, pruned_loss=0.06055, over 1615424.56 frames. ], batch size: 19, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:27:52,742 INFO [zipformer.py:1185] (0/4) 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] (0/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-07 07:28:19,019 INFO [optim.py:369] (0/4) 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 07:28:25,093 INFO [train.py:901] (0/4) Epoch 24, batch 5800, loss[loss=0.1936, simple_loss=0.2709, pruned_loss=0.05818, over 7233.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2856, pruned_loss=0.0604, over 1616290.07 frames. ], batch size: 16, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:28:38,084 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=191727.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:28:46,950 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2519, 2.4371, 2.0025, 2.9718, 1.4406, 1.8135, 2.1706, 2.2961], + device='cuda:0'), covar=tensor([0.0682, 0.0661, 0.0825, 0.0295, 0.0963, 0.1121, 0.0725, 0.0777], + device='cuda:0'), in_proj_covar=tensor([0.0235, 0.0199, 0.0246, 0.0216, 0.0206, 0.0247, 0.0253, 0.0209], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 07:28:59,633 INFO [train.py:901] (0/4) Epoch 24, batch 5850, loss[loss=0.2057, simple_loss=0.2947, pruned_loss=0.05834, over 8236.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2852, pruned_loss=0.06032, over 1613754.80 frames. ], batch size: 22, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:29:29,158 INFO [optim.py:369] (0/4) 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 07:29:30,137 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=191801.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:29:34,717 INFO [train.py:901] (0/4) Epoch 24, batch 5900, loss[loss=0.1868, simple_loss=0.2785, pruned_loss=0.04754, over 8036.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2858, pruned_loss=0.06019, over 1620010.15 frames. ], batch size: 22, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:29:45,873 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 +2023-02-07 07:29:48,205 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=191826.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:29:56,045 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 +2023-02-07 07:29:59,334 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191842.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:30:10,957 INFO [train.py:901] (0/4) Epoch 24, batch 5950, loss[loss=0.2362, simple_loss=0.313, pruned_loss=0.07965, over 8705.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2833, pruned_loss=0.05906, over 1614056.75 frames. ], batch size: 34, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:30:16,050 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=191865.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:30:33,606 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=191890.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:30:34,659 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 +2023-02-07 07:30:40,226 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 24, batch 6000, loss[loss=0.2412, simple_loss=0.3147, pruned_loss=0.08381, over 7204.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2846, pruned_loss=0.05951, over 1614776.65 frames. ], batch size: 71, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:30:45,682 INFO [train.py:926] (0/4) Computing validation loss +2023-02-07 07:31:01,024 INFO [train.py:935] (0/4) Epoch 24, validation: loss=0.1718, simple_loss=0.2718, pruned_loss=0.0359, over 944034.00 frames. +2023-02-07 07:31:01,026 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-07 07:31:08,198 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=191918.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:31:13,481 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.5078, 1.6381, 4.6952, 1.9845, 4.1909, 3.9157, 4.2996, 4.1472], + device='cuda:0'), covar=tensor([0.0583, 0.4870, 0.0502, 0.4147, 0.1059, 0.0969, 0.0558, 0.0637], + device='cuda:0'), in_proj_covar=tensor([0.0648, 0.0653, 0.0711, 0.0644, 0.0720, 0.0616, 0.0618, 0.0690], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 07:31:24,825 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=191943.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:31:35,325 INFO [train.py:901] (0/4) Epoch 24, batch 6050, loss[loss=0.2156, simple_loss=0.3052, pruned_loss=0.06297, over 8077.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2847, pruned_loss=0.05936, over 1618171.86 frames. ], batch size: 21, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:31:46,660 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([6.1433, 2.0375, 6.2533, 2.3723, 5.7465, 5.2529, 5.8407, 5.6975], + device='cuda:0'), covar=tensor([0.0348, 0.4412, 0.0277, 0.3714, 0.0733, 0.0787, 0.0401, 0.0420], + device='cuda:0'), in_proj_covar=tensor([0.0646, 0.0652, 0.0709, 0.0641, 0.0718, 0.0615, 0.0618, 0.0688], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 07:32:04,602 INFO [optim.py:369] (0/4) 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 07:32:04,770 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-192000.pt +2023-02-07 07:32:11,925 INFO [train.py:901] (0/4) Epoch 24, batch 6100, loss[loss=0.1695, simple_loss=0.2546, pruned_loss=0.04223, over 7657.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.284, pruned_loss=0.05929, over 1611703.37 frames. ], batch size: 19, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:32:15,176 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-02-07 07:32:25,157 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7456, 1.4670, 3.1792, 1.1876, 2.2850, 3.5117, 3.8330, 2.5738], + device='cuda:0'), covar=tensor([0.1729, 0.2455, 0.0513, 0.3225, 0.1318, 0.0402, 0.0653, 0.1125], + device='cuda:0'), in_proj_covar=tensor([0.0295, 0.0322, 0.0286, 0.0313, 0.0313, 0.0272, 0.0428, 0.0302], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 07:32:34,086 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-07 07:32:46,756 INFO [train.py:901] (0/4) Epoch 24, batch 6150, loss[loss=0.2107, simple_loss=0.2774, pruned_loss=0.07202, over 7225.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2838, pruned_loss=0.05928, over 1611488.16 frames. ], batch size: 16, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:33:15,568 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192098.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:33:16,714 INFO [optim.py:369] (0/4) 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 07:33:20,331 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 +2023-02-07 07:33:22,018 INFO [train.py:901] (0/4) Epoch 24, batch 6200, loss[loss=0.2036, simple_loss=0.2831, pruned_loss=0.06203, over 7977.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2841, pruned_loss=0.05971, over 1613776.35 frames. ], batch size: 21, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:33:30,657 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192120.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:33:32,760 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192123.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:33:56,607 INFO [train.py:901] (0/4) Epoch 24, batch 6250, loss[loss=0.208, simple_loss=0.2867, pruned_loss=0.06467, over 8285.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2836, pruned_loss=0.05909, over 1615207.99 frames. ], batch size: 23, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:34:07,851 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192173.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:34:26,658 INFO [optim.py:369] (0/4) 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 07:34:33,020 INFO [train.py:901] (0/4) Epoch 24, batch 6300, loss[loss=0.196, simple_loss=0.2664, pruned_loss=0.06277, over 7546.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2833, pruned_loss=0.05881, over 1611313.47 frames. ], batch size: 18, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:34:52,685 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192237.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:34:54,074 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192239.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:34:54,839 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7629, 1.7261, 2.6992, 2.0574, 2.4605, 1.7951, 1.5570, 1.2497], + device='cuda:0'), covar=tensor([0.7216, 0.6072, 0.2045, 0.3961, 0.2917, 0.4418, 0.3125, 0.5664], + device='cuda:0'), in_proj_covar=tensor([0.0949, 0.0998, 0.0817, 0.0970, 0.1005, 0.0910, 0.0760, 0.0833], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 07:35:07,361 INFO [train.py:901] (0/4) Epoch 24, batch 6350, loss[loss=0.2095, simple_loss=0.2933, pruned_loss=0.06285, over 8477.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2849, pruned_loss=0.05978, over 1614544.03 frames. ], batch size: 25, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:35:20,870 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 +2023-02-07 07:35:36,835 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 24, batch 6400, loss[loss=0.223, simple_loss=0.3084, pruned_loss=0.06878, over 8286.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.285, pruned_loss=0.05999, over 1619850.35 frames. ], batch size: 23, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:35:58,398 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7297, 1.4487, 2.8605, 1.4088, 2.3070, 3.0888, 3.2146, 2.6678], + device='cuda:0'), covar=tensor([0.1074, 0.1542, 0.0343, 0.1953, 0.0768, 0.0291, 0.0650, 0.0548], + device='cuda:0'), in_proj_covar=tensor([0.0297, 0.0323, 0.0286, 0.0315, 0.0314, 0.0272, 0.0430, 0.0303], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 07:36:19,118 INFO [train.py:901] (0/4) Epoch 24, batch 6450, loss[loss=0.1787, simple_loss=0.2402, pruned_loss=0.05863, over 7444.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2855, pruned_loss=0.06031, over 1619562.00 frames. ], batch size: 17, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:36:49,115 INFO [optim.py:369] (0/4) 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 07:36:53,706 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 +2023-02-07 07:36:54,692 INFO [train.py:901] (0/4) Epoch 24, batch 6500, loss[loss=0.1552, simple_loss=0.2408, pruned_loss=0.03477, over 7654.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2843, pruned_loss=0.05951, over 1614353.37 frames. ], batch size: 19, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:36:56,119 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192410.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:37:00,677 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192417.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:37:24,989 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192451.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:37:29,457 INFO [train.py:901] (0/4) Epoch 24, batch 6550, loss[loss=0.2488, simple_loss=0.3169, pruned_loss=0.09034, over 6642.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2855, pruned_loss=0.06031, over 1611856.72 frames. ], batch size: 71, lr: 3.13e-03, grad_scale: 8.0 +2023-02-07 07:37:33,652 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=192464.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:37:48,308 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-07 07:37:58,370 INFO [optim.py:369] (0/4) 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 07:38:03,874 INFO [train.py:901] (0/4) Epoch 24, batch 6600, loss[loss=0.22, simple_loss=0.3061, pruned_loss=0.06693, over 8616.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2845, pruned_loss=0.05982, over 1610728.53 frames. ], batch size: 39, lr: 3.12e-03, grad_scale: 8.0 +2023-02-07 07:38:08,128 WARNING [train.py:1067] (0/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-07 07:38:10,832 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=192517.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:38:18,041 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 +2023-02-07 07:38:38,394 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.6120, 1.6995, 3.7703, 1.5441, 3.3533, 3.1532, 3.4142, 3.2964], + device='cuda:0'), covar=tensor([0.0740, 0.3818, 0.0731, 0.4032, 0.1206, 0.0979, 0.0668, 0.0741], + device='cuda:0'), in_proj_covar=tensor([0.0646, 0.0655, 0.0710, 0.0639, 0.0718, 0.0614, 0.0617, 0.0686], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 07:38:39,631 INFO [train.py:901] (0/4) Epoch 24, batch 6650, loss[loss=0.2133, simple_loss=0.2955, pruned_loss=0.06551, over 8337.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2846, pruned_loss=0.0597, over 1614109.68 frames. ], batch size: 26, lr: 3.12e-03, grad_scale: 8.0 +2023-02-07 07:38:53,940 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=192581.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:38:56,512 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=192583.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:39:08,651 INFO [optim.py:369] (0/4) 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 07:39:14,186 INFO [train.py:901] (0/4) Epoch 24, batch 6700, loss[loss=0.1848, simple_loss=0.2823, pruned_loss=0.04364, over 8320.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2841, pruned_loss=0.0597, over 1614474.04 frames. ], batch size: 25, lr: 3.12e-03, grad_scale: 8.0 +2023-02-07 07:39:31,214 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=192632.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:39:32,530 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192634.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:39:49,031 INFO [train.py:901] (0/4) Epoch 24, batch 6750, loss[loss=0.2248, simple_loss=0.3032, pruned_loss=0.07319, over 7920.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2841, pruned_loss=0.05989, over 1608406.61 frames. ], batch size: 20, lr: 3.12e-03, grad_scale: 8.0 +2023-02-07 07:40:12,029 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192691.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:40:15,265 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=192696.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:40:16,610 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=192698.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:40:17,826 INFO [optim.py:369] (0/4) 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 07:40:23,409 INFO [train.py:901] (0/4) Epoch 24, batch 6800, loss[loss=0.2284, simple_loss=0.2981, pruned_loss=0.07937, over 7556.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2865, pruned_loss=0.06106, over 1611805.59 frames. ], batch size: 18, lr: 3.12e-03, grad_scale: 8.0 +2023-02-07 07:40:24,836 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-07 07:40:40,562 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.15 vs. limit=5.0 +2023-02-07 07:40:56,186 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192753.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:40:56,807 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 24, batch 6850, loss[loss=0.1982, simple_loss=0.2921, pruned_loss=0.0522, over 8324.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2858, pruned_loss=0.0604, over 1614485.93 frames. ], batch size: 25, lr: 3.12e-03, grad_scale: 8.0 +2023-02-07 07:41:01,648 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=192761.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:41:15,307 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-07 07:41:26,163 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=192795.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:41:27,525 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3302, 1.4092, 4.1708, 2.0091, 2.4542, 4.8804, 5.0076, 4.2102], + device='cuda:0'), covar=tensor([0.1072, 0.2211, 0.0320, 0.1956, 0.1330, 0.0181, 0.0369, 0.0524], + device='cuda:0'), in_proj_covar=tensor([0.0298, 0.0325, 0.0287, 0.0315, 0.0315, 0.0273, 0.0431, 0.0304], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 07:41:29,431 INFO [optim.py:369] (0/4) 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 07:41:34,274 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 +2023-02-07 07:41:35,208 INFO [train.py:901] (0/4) Epoch 24, batch 6900, loss[loss=0.2752, simple_loss=0.3389, pruned_loss=0.1058, over 7229.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2856, pruned_loss=0.06053, over 1609468.86 frames. ], batch size: 71, lr: 3.12e-03, grad_scale: 8.0 +2023-02-07 07:41:48,734 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5147, 2.0107, 3.0785, 1.3952, 2.2960, 1.9984, 1.6900, 2.4344], + device='cuda:0'), covar=tensor([0.2129, 0.2711, 0.1025, 0.4946, 0.2113, 0.3336, 0.2582, 0.2407], + device='cuda:0'), in_proj_covar=tensor([0.0532, 0.0619, 0.0557, 0.0655, 0.0652, 0.0601, 0.0550, 0.0639], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 07:41:54,061 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192835.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:42:03,101 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192849.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 07:42:08,868 INFO [train.py:901] (0/4) Epoch 24, batch 6950, loss[loss=0.2116, simple_loss=0.2989, pruned_loss=0.06213, over 8531.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2855, pruned_loss=0.06017, over 1609819.04 frames. ], batch size: 28, lr: 3.12e-03, grad_scale: 8.0 +2023-02-07 07:42:10,319 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192860.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:42:17,130 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=192869.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:42:21,962 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=192876.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:42:23,178 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-07 07:42:30,036 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192888.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:42:37,505 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6172, 2.0175, 3.1696, 1.4944, 2.4593, 2.1260, 1.7240, 2.5109], + device='cuda:0'), covar=tensor([0.1958, 0.2730, 0.0957, 0.4747, 0.1918, 0.3214, 0.2436, 0.2240], + device='cuda:0'), in_proj_covar=tensor([0.0533, 0.0620, 0.0558, 0.0656, 0.0653, 0.0602, 0.0550, 0.0640], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 07:42:38,586 INFO [optim.py:369] (0/4) 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 07:42:44,708 INFO [train.py:901] (0/4) Epoch 24, batch 7000, loss[loss=0.2362, simple_loss=0.3176, pruned_loss=0.07737, over 8453.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2853, pruned_loss=0.05973, over 1608919.81 frames. ], batch size: 29, lr: 3.12e-03, grad_scale: 8.0 +2023-02-07 07:42:46,232 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=192910.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:42:48,260 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192913.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:43:15,560 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192952.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:43:16,882 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192954.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:43:19,416 INFO [train.py:901] (0/4) Epoch 24, batch 7050, loss[loss=0.2368, simple_loss=0.322, pruned_loss=0.07579, over 8692.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2864, pruned_loss=0.0603, over 1609925.56 frames. ], batch size: 39, lr: 3.12e-03, grad_scale: 8.0 +2023-02-07 07:43:30,606 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3287, 2.1294, 2.7273, 2.2710, 2.7138, 2.3714, 2.1639, 1.5370], + device='cuda:0'), covar=tensor([0.5360, 0.4903, 0.2034, 0.3694, 0.2636, 0.3071, 0.1902, 0.5209], + device='cuda:0'), in_proj_covar=tensor([0.0945, 0.0995, 0.0815, 0.0964, 0.1001, 0.0907, 0.0756, 0.0830], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 07:43:32,666 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192977.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:43:33,239 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=192978.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:43:34,092 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192979.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:43:48,988 INFO [optim.py:369] (0/4) 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 07:43:54,456 INFO [train.py:901] (0/4) Epoch 24, batch 7100, loss[loss=0.1768, simple_loss=0.2608, pruned_loss=0.04639, over 7802.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2843, pruned_loss=0.05941, over 1609201.15 frames. ], batch size: 19, lr: 3.12e-03, grad_scale: 8.0 +2023-02-07 07:44:14,194 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=193035.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:44:29,512 INFO [train.py:901] (0/4) Epoch 24, batch 7150, loss[loss=0.1977, simple_loss=0.2676, pruned_loss=0.06385, over 7521.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.283, pruned_loss=0.05897, over 1603932.11 frames. ], batch size: 18, lr: 3.12e-03, grad_scale: 8.0 +2023-02-07 07:44:30,377 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193059.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:44:54,302 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193093.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:44:55,772 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7923, 1.6668, 2.6131, 2.0371, 2.3148, 1.7986, 1.5258, 1.1731], + device='cuda:0'), covar=tensor([0.8014, 0.7437, 0.2342, 0.4315, 0.3392, 0.5025, 0.3568, 0.6316], + device='cuda:0'), in_proj_covar=tensor([0.0948, 0.0997, 0.0817, 0.0965, 0.1006, 0.0910, 0.0759, 0.0832], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 07:44:56,926 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=193097.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:44:58,925 INFO [optim.py:369] (0/4) 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 07:45:05,029 INFO [train.py:901] (0/4) Epoch 24, batch 7200, loss[loss=0.2321, simple_loss=0.3049, pruned_loss=0.07968, over 8656.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2844, pruned_loss=0.05938, over 1609571.69 frames. ], batch size: 39, lr: 3.12e-03, grad_scale: 8.0 +2023-02-07 07:45:16,877 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193125.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:45:21,521 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193132.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:45:22,767 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1999, 1.7234, 4.4615, 2.0728, 2.5699, 5.0433, 5.1235, 4.2304], + device='cuda:0'), covar=tensor([0.1346, 0.1967, 0.0309, 0.1998, 0.1139, 0.0214, 0.0540, 0.0649], + device='cuda:0'), in_proj_covar=tensor([0.0296, 0.0321, 0.0286, 0.0313, 0.0313, 0.0272, 0.0428, 0.0302], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 07:45:34,268 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193150.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:45:34,292 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193157.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:45:40,076 INFO [train.py:901] (0/4) Epoch 24, batch 7250, loss[loss=0.1962, simple_loss=0.2815, pruned_loss=0.05545, over 8554.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2838, pruned_loss=0.05925, over 1612805.86 frames. ], batch size: 39, lr: 3.12e-03, grad_scale: 16.0 +2023-02-07 07:45:45,711 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193166.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:46:03,183 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193191.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:46:04,338 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) Epoch 24, batch 7300, loss[loss=0.1722, simple_loss=0.2481, pruned_loss=0.04815, over 7798.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2836, pruned_loss=0.059, over 1614519.01 frames. ], batch size: 19, lr: 3.12e-03, grad_scale: 16.0 +2023-02-07 07:46:17,189 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193212.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:46:29,138 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193229.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:46:41,117 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6343, 1.3721, 2.9053, 1.4849, 2.2377, 3.0757, 3.2172, 2.6577], + device='cuda:0'), covar=tensor([0.1148, 0.1716, 0.0364, 0.1940, 0.0889, 0.0315, 0.0578, 0.0565], + device='cuda:0'), in_proj_covar=tensor([0.0295, 0.0320, 0.0284, 0.0311, 0.0312, 0.0270, 0.0426, 0.0300], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 07:46:48,079 INFO [train.py:901] (0/4) Epoch 24, batch 7350, loss[loss=0.1938, simple_loss=0.2898, pruned_loss=0.04885, over 8497.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2824, pruned_loss=0.05885, over 1611238.20 frames. ], batch size: 29, lr: 3.12e-03, grad_scale: 16.0 +2023-02-07 07:47:11,633 WARNING [train.py:1067] (0/4) 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] (0/4) 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 07:47:19,304 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193302.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:47:23,205 INFO [train.py:901] (0/4) Epoch 24, batch 7400, loss[loss=0.1565, simple_loss=0.2453, pruned_loss=0.03382, over 7655.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2834, pruned_loss=0.05938, over 1608777.27 frames. ], batch size: 19, lr: 3.12e-03, grad_scale: 16.0 +2023-02-07 07:47:23,395 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193308.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 07:47:31,916 WARNING [train.py:1067] (0/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-07 07:47:51,095 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5625, 1.9488, 3.0402, 1.4445, 2.2284, 2.0102, 1.6386, 2.3231], + device='cuda:0'), covar=tensor([0.1998, 0.2707, 0.0942, 0.4761, 0.2112, 0.3232, 0.2409, 0.2274], + device='cuda:0'), in_proj_covar=tensor([0.0528, 0.0616, 0.0555, 0.0649, 0.0648, 0.0597, 0.0545, 0.0635], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 07:47:52,375 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 24, batch 7450, loss[loss=0.2023, simple_loss=0.2998, pruned_loss=0.05241, over 8301.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2835, pruned_loss=0.05899, over 1609962.48 frames. ], batch size: 23, lr: 3.12e-03, grad_scale: 16.0 +2023-02-07 07:48:09,518 WARNING [train.py:1067] (0/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-07 07:48:10,314 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193374.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:48:24,526 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1005, 2.2655, 1.8786, 2.8231, 1.2415, 1.7359, 2.0121, 2.2497], + device='cuda:0'), covar=tensor([0.0683, 0.0707, 0.0869, 0.0317, 0.1116, 0.1162, 0.0775, 0.0797], + device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0198, 0.0245, 0.0214, 0.0206, 0.0248, 0.0252, 0.0208], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 07:48:28,444 INFO [optim.py:369] (0/4) 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 07:48:30,628 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=193403.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:48:32,862 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 24, batch 7500, loss[loss=0.1731, simple_loss=0.2666, pruned_loss=0.03981, over 8249.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2852, pruned_loss=0.0597, over 1613605.16 frames. ], batch size: 24, lr: 3.12e-03, grad_scale: 16.0 +2023-02-07 07:48:44,182 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.20 vs. limit=5.0 +2023-02-07 07:48:50,681 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) Epoch 24, batch 7550, loss[loss=0.1952, simple_loss=0.282, pruned_loss=0.05415, over 8229.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.285, pruned_loss=0.05944, over 1616173.78 frames. ], batch size: 22, lr: 3.12e-03, grad_scale: 16.0 +2023-02-07 07:49:16,842 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193468.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:49:19,403 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193472.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 07:49:34,380 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193493.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:49:34,556 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 +2023-02-07 07:49:39,065 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 24, batch 7600, loss[loss=0.217, simple_loss=0.3037, pruned_loss=0.06519, over 8554.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2842, pruned_loss=0.05942, over 1614487.90 frames. ], batch size: 31, lr: 3.12e-03, grad_scale: 16.0 +2023-02-07 07:49:51,992 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 24, batch 7650, loss[loss=0.2027, simple_loss=0.2866, pruned_loss=0.05938, over 8359.00 frames. ], tot_loss[loss=0.203, simple_loss=0.286, pruned_loss=0.05999, over 1618694.84 frames. ], batch size: 24, lr: 3.12e-03, grad_scale: 16.0 +2023-02-07 07:50:23,598 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=193573.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:50:41,146 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193589.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 07:50:48,583 INFO [optim.py:369] (0/4) 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 07:50:53,968 INFO [train.py:901] (0/4) Epoch 24, batch 7700, loss[loss=0.2207, simple_loss=0.2884, pruned_loss=0.07656, over 7982.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2871, pruned_loss=0.06062, over 1619589.11 frames. ], batch size: 21, lr: 3.12e-03, grad_scale: 16.0 +2023-02-07 07:51:11,794 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193633.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:51:14,901 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-07 07:51:20,922 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=193646.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:51:29,067 INFO [train.py:901] (0/4) Epoch 24, batch 7750, loss[loss=0.1941, simple_loss=0.2742, pruned_loss=0.05703, over 7250.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2862, pruned_loss=0.06012, over 1616824.89 frames. ], batch size: 16, lr: 3.12e-03, grad_scale: 16.0 +2023-02-07 07:51:50,176 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193688.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:51:51,544 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9275, 1.7419, 2.5765, 1.6743, 1.4002, 2.4892, 0.4618, 1.6220], + device='cuda:0'), covar=tensor([0.1444, 0.1316, 0.0290, 0.1162, 0.2469, 0.0322, 0.2018, 0.1154], + device='cuda:0'), in_proj_covar=tensor([0.0196, 0.0202, 0.0130, 0.0221, 0.0274, 0.0140, 0.0172, 0.0198], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 07:51:58,105 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 24, batch 7800, loss[loss=0.2573, simple_loss=0.3392, pruned_loss=0.08773, over 8129.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2856, pruned_loss=0.06003, over 1613457.07 frames. ], batch size: 22, lr: 3.12e-03, grad_scale: 16.0 +2023-02-07 07:52:10,492 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3741, 1.3634, 2.3238, 1.1966, 2.2573, 2.4614, 2.6314, 1.9917], + device='cuda:0'), covar=tensor([0.1290, 0.1522, 0.0507, 0.2337, 0.0744, 0.0484, 0.0814, 0.0899], + device='cuda:0'), in_proj_covar=tensor([0.0299, 0.0322, 0.0287, 0.0316, 0.0315, 0.0272, 0.0429, 0.0302], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 07:52:37,266 INFO [train.py:901] (0/4) Epoch 24, batch 7850, loss[loss=0.1887, simple_loss=0.2673, pruned_loss=0.05508, over 7233.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2858, pruned_loss=0.06058, over 1609709.51 frames. ], batch size: 16, lr: 3.11e-03, grad_scale: 16.0 +2023-02-07 07:52:39,490 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193761.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:52:48,223 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193774.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:52:54,176 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=193783.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:53:05,048 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) Epoch 24, batch 7900, loss[loss=0.2376, simple_loss=0.3117, pruned_loss=0.08178, over 7096.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.287, pruned_loss=0.06085, over 1613644.47 frames. ], batch size: 71, lr: 3.11e-03, grad_scale: 16.0 +2023-02-07 07:53:16,263 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=193816.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 07:53:43,830 INFO [train.py:901] (0/4) Epoch 24, batch 7950, loss[loss=0.2147, simple_loss=0.2995, pruned_loss=0.06499, over 8589.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2867, pruned_loss=0.0606, over 1614150.21 frames. ], batch size: 31, lr: 3.11e-03, grad_scale: 16.0 +2023-02-07 07:54:11,276 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 07:54:13,962 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193902.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:54:17,820 INFO [train.py:901] (0/4) Epoch 24, batch 8000, loss[loss=0.1686, simple_loss=0.2463, pruned_loss=0.04539, over 7421.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2882, pruned_loss=0.06114, over 1613919.03 frames. ], batch size: 17, lr: 3.11e-03, grad_scale: 16.0 +2023-02-07 07:54:33,377 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193931.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 07:54:42,121 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193944.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:54:51,054 INFO [train.py:901] (0/4) Epoch 24, batch 8050, loss[loss=0.1917, simple_loss=0.2649, pruned_loss=0.05926, over 7439.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2851, pruned_loss=0.06049, over 1592752.82 frames. ], batch size: 17, lr: 3.11e-03, grad_scale: 16.0 +2023-02-07 07:54:58,121 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193969.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:55:03,435 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=193977.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:55:13,098 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/epoch-24.pt +2023-02-07 07:55:25,187 WARNING [train.py:1067] (0/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-07 07:55:28,445 INFO [train.py:901] (0/4) Epoch 25, batch 0, loss[loss=0.2769, simple_loss=0.3367, pruned_loss=0.1086, over 7815.00 frames. ], tot_loss[loss=0.2769, simple_loss=0.3367, pruned_loss=0.1086, over 7815.00 frames. ], batch size: 20, lr: 3.05e-03, grad_scale: 16.0 +2023-02-07 07:55:28,446 INFO [train.py:926] (0/4) Computing validation loss +2023-02-07 07:55:39,669 INFO [train.py:935] (0/4) Epoch 25, validation: loss=0.1722, simple_loss=0.2724, pruned_loss=0.03604, over 944034.00 frames. +2023-02-07 07:55:39,670 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-07 07:55:46,485 INFO [optim.py:369] (0/4) 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 07:55:46,615 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-194000.pt +2023-02-07 07:55:57,050 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-07 07:56:00,103 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194017.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:56:15,973 INFO [train.py:901] (0/4) Epoch 25, batch 50, loss[loss=0.2154, simple_loss=0.2813, pruned_loss=0.07476, over 5576.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2858, pruned_loss=0.0605, over 363480.64 frames. ], batch size: 12, lr: 3.05e-03, grad_scale: 16.0 +2023-02-07 07:56:17,571 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194042.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:56:32,512 WARNING [train.py:1067] (0/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-07 07:56:51,125 INFO [train.py:901] (0/4) Epoch 25, batch 100, loss[loss=0.2202, simple_loss=0.3073, pruned_loss=0.06659, over 8295.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2901, pruned_loss=0.06273, over 647972.67 frames. ], batch size: 23, lr: 3.05e-03, grad_scale: 16.0 +2023-02-07 07:56:51,259 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194090.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:56:52,651 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194092.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:56:55,677 WARNING [train.py:1067] (0/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-07 07:56:57,725 INFO [optim.py:369] (0/4) 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 07:57:02,239 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 +2023-02-07 07:57:11,416 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.1867, 1.0646, 1.2831, 1.0316, 0.9482, 1.3188, 0.0845, 1.0021], + device='cuda:0'), covar=tensor([0.1504, 0.1210, 0.0436, 0.0695, 0.2554, 0.0514, 0.1973, 0.1042], + device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0199, 0.0129, 0.0218, 0.0271, 0.0138, 0.0171, 0.0196], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 07:57:13,418 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7352, 1.8859, 1.5871, 2.3264, 0.9956, 1.4388, 1.6663, 1.8868], + device='cuda:0'), covar=tensor([0.0754, 0.0734, 0.0921, 0.0369, 0.1093, 0.1372, 0.0748, 0.0733], + device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0197, 0.0244, 0.0214, 0.0205, 0.0247, 0.0251, 0.0206], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 07:57:25,367 INFO [train.py:901] (0/4) Epoch 25, batch 150, loss[loss=0.2023, simple_loss=0.2951, pruned_loss=0.05475, over 8776.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2889, pruned_loss=0.06185, over 865086.86 frames. ], batch size: 30, lr: 3.05e-03, grad_scale: 16.0 +2023-02-07 07:57:35,091 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194154.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:57:43,933 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5368, 1.3464, 1.5938, 1.3311, 0.8656, 1.3492, 1.4296, 1.2536], + device='cuda:0'), covar=tensor([0.0662, 0.1294, 0.1618, 0.1482, 0.0655, 0.1535, 0.0781, 0.0705], + device='cuda:0'), in_proj_covar=tensor([0.0098, 0.0152, 0.0188, 0.0159, 0.0100, 0.0162, 0.0112, 0.0144], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0009, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:0') +2023-02-07 07:57:52,148 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194179.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:57:58,201 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194187.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 07:58:00,657 INFO [train.py:901] (0/4) Epoch 25, batch 200, loss[loss=0.1907, simple_loss=0.2626, pruned_loss=0.05939, over 7419.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2872, pruned_loss=0.06126, over 1031067.38 frames. ], batch size: 17, lr: 3.05e-03, grad_scale: 16.0 +2023-02-07 07:58:04,983 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5253, 1.9083, 2.9653, 1.3980, 2.1699, 1.8768, 1.6312, 2.1744], + device='cuda:0'), covar=tensor([0.2025, 0.2702, 0.0849, 0.4762, 0.1968, 0.3460, 0.2521, 0.2283], + device='cuda:0'), in_proj_covar=tensor([0.0526, 0.0615, 0.0554, 0.0651, 0.0649, 0.0596, 0.0546, 0.0633], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 07:58:07,397 INFO [optim.py:369] (0/4) 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 07:58:14,561 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4696, 1.3418, 1.7014, 1.2668, 0.9062, 1.4228, 1.4364, 1.2633], + device='cuda:0'), covar=tensor([0.0660, 0.1335, 0.1709, 0.1554, 0.0626, 0.1554, 0.0735, 0.0703], + device='cuda:0'), in_proj_covar=tensor([0.0098, 0.0153, 0.0189, 0.0159, 0.0101, 0.0163, 0.0112, 0.0145], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:0') +2023-02-07 07:58:16,685 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 25, batch 250, loss[loss=0.2184, simple_loss=0.3021, pruned_loss=0.06735, over 8506.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2855, pruned_loss=0.05925, over 1164104.29 frames. ], batch size: 28, lr: 3.05e-03, grad_scale: 16.0 +2023-02-07 07:58:39,443 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=194246.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:58:49,470 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-07 07:58:56,324 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6875, 2.0091, 2.1424, 1.3120, 2.2188, 1.5472, 0.6425, 1.8507], + device='cuda:0'), covar=tensor([0.0610, 0.0366, 0.0308, 0.0642, 0.0394, 0.0903, 0.0985, 0.0374], + device='cuda:0'), in_proj_covar=tensor([0.0461, 0.0400, 0.0356, 0.0452, 0.0386, 0.0539, 0.0398, 0.0431], + device='cuda:0'), out_proj_covar=tensor([1.2274e-04, 1.0453e-04, 9.3081e-05, 1.1866e-04, 1.0132e-04, 1.5105e-04, + 1.0688e-04, 1.1364e-04], device='cuda:0') +2023-02-07 07:58:58,133 WARNING [train.py:1067] (0/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-07 07:59:09,575 INFO [train.py:901] (0/4) Epoch 25, batch 300, loss[loss=0.1739, simple_loss=0.2682, pruned_loss=0.03976, over 8338.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.284, pruned_loss=0.05851, over 1267494.39 frames. ], batch size: 25, lr: 3.05e-03, grad_scale: 16.0 +2023-02-07 07:59:17,096 INFO [optim.py:369] (0/4) 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 07:59:35,717 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 +2023-02-07 07:59:43,417 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194336.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:59:45,916 INFO [train.py:901] (0/4) Epoch 25, batch 350, loss[loss=0.2137, simple_loss=0.2999, pruned_loss=0.06377, over 8264.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2855, pruned_loss=0.05917, over 1346770.04 frames. ], batch size: 24, lr: 3.05e-03, grad_scale: 16.0 +2023-02-07 07:59:51,517 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194348.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 07:59:55,579 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.8893, 2.0567, 4.0350, 2.2913, 3.6635, 3.4325, 3.7665, 3.6657], + device='cuda:0'), covar=tensor([0.0733, 0.3699, 0.0846, 0.3898, 0.1095, 0.0904, 0.0617, 0.0676], + device='cuda:0'), in_proj_covar=tensor([0.0645, 0.0655, 0.0710, 0.0646, 0.0717, 0.0612, 0.0617, 0.0692], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 08:00:00,424 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194361.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:00:08,024 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194371.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:00:09,421 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194373.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:00:14,123 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194380.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 08:00:20,644 INFO [train.py:901] (0/4) Epoch 25, batch 400, loss[loss=0.23, simple_loss=0.3129, pruned_loss=0.07355, over 8341.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2854, pruned_loss=0.05971, over 1403703.33 frames. ], batch size: 26, lr: 3.05e-03, grad_scale: 16.0 +2023-02-07 08:00:27,611 INFO [optim.py:369] (0/4) 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 08:00:27,884 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.7369, 1.7052, 5.8784, 2.1154, 5.2957, 4.9390, 5.4319, 5.2861], + device='cuda:0'), covar=tensor([0.0435, 0.4753, 0.0397, 0.4171, 0.0930, 0.0796, 0.0458, 0.0509], + device='cuda:0'), in_proj_covar=tensor([0.0645, 0.0655, 0.0709, 0.0644, 0.0715, 0.0611, 0.0616, 0.0690], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 08:00:52,190 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=194434.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:00:56,957 INFO [train.py:901] (0/4) Epoch 25, batch 450, loss[loss=0.195, simple_loss=0.2789, pruned_loss=0.05557, over 8105.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2857, pruned_loss=0.05903, over 1454155.99 frames. ], batch size: 23, lr: 3.05e-03, grad_scale: 16.0 +2023-02-07 08:00:59,297 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8271, 1.8669, 1.6865, 2.3408, 0.9844, 1.6163, 1.6946, 1.9217], + device='cuda:0'), covar=tensor([0.0761, 0.0763, 0.0925, 0.0400, 0.1050, 0.1181, 0.0736, 0.0738], + device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0197, 0.0245, 0.0214, 0.0205, 0.0247, 0.0251, 0.0207], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 08:01:07,926 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4239, 2.6511, 3.0319, 1.8338, 3.3180, 2.1387, 1.6168, 2.3925], + device='cuda:0'), covar=tensor([0.0718, 0.0403, 0.0277, 0.0760, 0.0417, 0.0819, 0.0903, 0.0489], + device='cuda:0'), in_proj_covar=tensor([0.0464, 0.0404, 0.0358, 0.0454, 0.0390, 0.0541, 0.0401, 0.0435], + device='cuda:0'), out_proj_covar=tensor([1.2337e-04, 1.0538e-04, 9.3850e-05, 1.1925e-04, 1.0229e-04, 1.5171e-04, + 1.0760e-04, 1.1450e-04], device='cuda:0') +2023-02-07 08:01:30,916 INFO [train.py:901] (0/4) Epoch 25, batch 500, loss[loss=0.22, simple_loss=0.2893, pruned_loss=0.07532, over 7680.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2859, pruned_loss=0.0602, over 1488714.42 frames. ], batch size: 18, lr: 3.04e-03, grad_scale: 16.0 +2023-02-07 08:01:37,833 INFO [optim.py:369] (0/4) 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 08:02:06,154 INFO [train.py:901] (0/4) Epoch 25, batch 550, loss[loss=0.1976, simple_loss=0.2917, pruned_loss=0.05181, over 8255.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2858, pruned_loss=0.06024, over 1518001.03 frames. ], batch size: 24, lr: 3.04e-03, grad_scale: 16.0 +2023-02-07 08:02:13,462 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194549.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:02:30,085 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0347, 1.4054, 1.5936, 1.4509, 1.0580, 1.3726, 1.9019, 1.9576], + device='cuda:0'), covar=tensor([0.0542, 0.1607, 0.2191, 0.1669, 0.0661, 0.1891, 0.0729, 0.0582], + device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0152, 0.0188, 0.0159, 0.0100, 0.0162, 0.0111, 0.0144], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:0') +2023-02-07 08:02:42,118 INFO [train.py:901] (0/4) Epoch 25, batch 600, loss[loss=0.1737, simple_loss=0.2526, pruned_loss=0.04741, over 6773.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2842, pruned_loss=0.05911, over 1537307.25 frames. ], batch size: 15, lr: 3.04e-03, grad_scale: 16.0 +2023-02-07 08:02:48,728 INFO [optim.py:369] (0/4) 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 08:03:01,134 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-07 08:03:01,346 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194617.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:03:16,772 INFO [train.py:901] (0/4) Epoch 25, batch 650, loss[loss=0.1986, simple_loss=0.2775, pruned_loss=0.05983, over 8290.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2838, pruned_loss=0.05868, over 1553489.09 frames. ], batch size: 23, lr: 3.04e-03, grad_scale: 8.0 +2023-02-07 08:03:18,060 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194642.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:03:25,629 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194652.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:03:45,798 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=194680.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:03:52,552 INFO [train.py:901] (0/4) Epoch 25, batch 700, loss[loss=0.1816, simple_loss=0.2539, pruned_loss=0.05465, over 7538.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2844, pruned_loss=0.0588, over 1568060.45 frames. ], batch size: 18, lr: 3.04e-03, grad_scale: 8.0 +2023-02-07 08:04:00,045 INFO [optim.py:369] (0/4) 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 08:04:09,639 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=194715.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:04:16,389 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=194724.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 08:04:27,392 INFO [train.py:901] (0/4) Epoch 25, batch 750, loss[loss=0.196, simple_loss=0.2838, pruned_loss=0.05411, over 8322.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2846, pruned_loss=0.05885, over 1582289.85 frames. ], batch size: 25, lr: 3.04e-03, grad_scale: 8.0 +2023-02-07 08:04:49,452 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-07 08:04:58,538 WARNING [train.py:1067] (0/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-07 08:05:03,369 INFO [train.py:901] (0/4) Epoch 25, batch 800, loss[loss=0.1925, simple_loss=0.2694, pruned_loss=0.05776, over 8132.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2839, pruned_loss=0.05886, over 1587474.28 frames. ], batch size: 22, lr: 3.04e-03, grad_scale: 8.0 +2023-02-07 08:05:07,611 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194795.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:05:11,583 INFO [optim.py:369] (0/4) 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 08:05:13,774 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1545, 2.2686, 1.9199, 2.7038, 1.3699, 1.8350, 2.0366, 2.4041], + device='cuda:0'), covar=tensor([0.0635, 0.0688, 0.0836, 0.0391, 0.1062, 0.1069, 0.0801, 0.0591], + device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0195, 0.0244, 0.0213, 0.0205, 0.0246, 0.0250, 0.0206], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 08:05:14,511 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194805.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:05:31,781 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194830.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:05:31,805 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194830.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:05:33,417 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.64 vs. limit=5.0 +2023-02-07 08:05:37,852 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194839.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 08:05:38,295 INFO [train.py:901] (0/4) Epoch 25, batch 850, loss[loss=0.2296, simple_loss=0.3067, pruned_loss=0.0762, over 8031.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2846, pruned_loss=0.0592, over 1596658.11 frames. ], batch size: 22, lr: 3.04e-03, grad_scale: 8.0 +2023-02-07 08:05:44,136 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4171, 1.4409, 1.4175, 1.8254, 0.7742, 1.3152, 1.4132, 1.5141], + device='cuda:0'), covar=tensor([0.0832, 0.0727, 0.0963, 0.0450, 0.1000, 0.1176, 0.0627, 0.0599], + device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0196, 0.0245, 0.0213, 0.0205, 0.0246, 0.0251, 0.0206], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 08:05:45,503 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2984, 2.6503, 2.0985, 3.8638, 1.5034, 1.9489, 2.3222, 2.8371], + device='cuda:0'), covar=tensor([0.0711, 0.0739, 0.0829, 0.0238, 0.1082, 0.1165, 0.0968, 0.0673], + device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0196, 0.0245, 0.0213, 0.0205, 0.0246, 0.0250, 0.0206], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 08:05:56,602 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194865.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:06:14,016 INFO [train.py:901] (0/4) Epoch 25, batch 900, loss[loss=0.2678, simple_loss=0.3461, pruned_loss=0.09472, over 8468.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2851, pruned_loss=0.05907, over 1601803.26 frames. ], batch size: 29, lr: 3.04e-03, grad_scale: 8.0 +2023-02-07 08:06:22,179 INFO [optim.py:369] (0/4) 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 08:06:28,887 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 +2023-02-07 08:06:44,145 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.38 vs. limit=5.0 +2023-02-07 08:06:49,715 INFO [train.py:901] (0/4) Epoch 25, batch 950, loss[loss=0.1847, simple_loss=0.2671, pruned_loss=0.05111, over 8083.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2858, pruned_loss=0.05977, over 1607339.35 frames. ], batch size: 21, lr: 3.04e-03, grad_scale: 8.0 +2023-02-07 08:07:13,450 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4921, 1.4729, 2.0519, 1.2788, 1.1025, 2.0287, 0.3427, 1.2775], + device='cuda:0'), covar=tensor([0.1723, 0.1372, 0.0366, 0.1163, 0.2829, 0.0417, 0.1977, 0.1316], + device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0199, 0.0129, 0.0219, 0.0270, 0.0137, 0.0170, 0.0194], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 08:07:17,973 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 +2023-02-07 08:07:19,514 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 25, batch 1000, loss[loss=0.1943, simple_loss=0.2737, pruned_loss=0.05746, over 7957.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2851, pruned_loss=0.05894, over 1612245.23 frames. ], batch size: 21, lr: 3.04e-03, grad_scale: 8.0 +2023-02-07 08:07:29,061 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=194996.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:07:32,165 INFO [optim.py:369] (0/4) 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 08:07:47,059 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9279, 1.4943, 3.4796, 1.4471, 2.3624, 3.7630, 3.9238, 3.2975], + device='cuda:0'), covar=tensor([0.1183, 0.1941, 0.0302, 0.2145, 0.1072, 0.0248, 0.0444, 0.0516], + device='cuda:0'), in_proj_covar=tensor([0.0299, 0.0323, 0.0287, 0.0317, 0.0315, 0.0274, 0.0430, 0.0303], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 08:07:51,774 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=195029.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:07:54,443 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-07 08:07:59,734 INFO [train.py:901] (0/4) Epoch 25, batch 1050, loss[loss=0.1881, simple_loss=0.2767, pruned_loss=0.04977, over 8283.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.285, pruned_loss=0.05885, over 1612700.49 frames. ], batch size: 23, lr: 3.04e-03, grad_scale: 8.0 +2023-02-07 08:08:06,388 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-07 08:08:07,151 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=195062.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 08:08:23,800 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=195076.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:08:31,306 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=195086.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:08:33,293 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=195089.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:08:33,843 INFO [train.py:901] (0/4) Epoch 25, batch 1100, loss[loss=0.1753, simple_loss=0.2575, pruned_loss=0.04659, over 8075.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.285, pruned_loss=0.0591, over 1613385.10 frames. ], batch size: 21, lr: 3.04e-03, grad_scale: 8.0 +2023-02-07 08:08:37,357 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=195095.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 08:08:41,229 INFO [optim.py:369] (0/4) 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 08:08:48,157 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195111.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:08:48,180 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=195111.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:08:54,919 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 25, batch 1150, loss[loss=0.1892, simple_loss=0.2687, pruned_loss=0.05489, over 7648.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2845, pruned_loss=0.05926, over 1610530.43 frames. ], batch size: 19, lr: 3.04e-03, grad_scale: 8.0 +2023-02-07 08:09:11,484 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.8705, 2.2999, 3.7307, 1.9423, 1.8941, 3.7359, 0.6316, 2.3034], + device='cuda:0'), covar=tensor([0.1434, 0.1262, 0.0248, 0.1736, 0.2351, 0.0250, 0.2102, 0.1252], + device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0199, 0.0129, 0.0220, 0.0271, 0.0138, 0.0170, 0.0195], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 08:09:16,871 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-07 08:09:43,693 INFO [train.py:901] (0/4) Epoch 25, batch 1200, loss[loss=0.1651, simple_loss=0.2402, pruned_loss=0.04499, over 7213.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2837, pruned_loss=0.0592, over 1603153.74 frames. ], batch size: 16, lr: 3.04e-03, grad_scale: 8.0 +2023-02-07 08:09:51,821 INFO [optim.py:369] (0/4) 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 08:09:57,121 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=195209.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:10:18,382 INFO [train.py:901] (0/4) Epoch 25, batch 1250, loss[loss=0.1781, simple_loss=0.2781, pruned_loss=0.03902, over 8033.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2847, pruned_loss=0.05946, over 1608720.57 frames. ], batch size: 22, lr: 3.04e-03, grad_scale: 8.0 +2023-02-07 08:10:20,610 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9965, 2.2790, 1.8330, 2.7779, 1.2892, 1.6672, 2.0438, 2.2149], + device='cuda:0'), covar=tensor([0.0662, 0.0634, 0.0873, 0.0336, 0.1066, 0.1215, 0.0734, 0.0729], + device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0197, 0.0246, 0.0214, 0.0206, 0.0247, 0.0251, 0.0208], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 08:10:53,105 INFO [train.py:901] (0/4) Epoch 25, batch 1300, loss[loss=0.217, simple_loss=0.3043, pruned_loss=0.06489, over 8466.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2844, pruned_loss=0.05899, over 1610616.70 frames. ], batch size: 25, lr: 3.04e-03, grad_scale: 8.0 +2023-02-07 08:11:00,277 INFO [optim.py:369] (0/4) 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 08:11:16,288 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195324.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:11:26,821 INFO [train.py:901] (0/4) Epoch 25, batch 1350, loss[loss=0.2291, simple_loss=0.3026, pruned_loss=0.07783, over 7769.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2842, pruned_loss=0.0593, over 1614374.45 frames. ], batch size: 19, lr: 3.04e-03, grad_scale: 8.0 +2023-02-07 08:11:42,194 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3631, 2.1332, 3.4775, 2.1021, 2.8546, 3.9191, 3.8833, 3.3856], + device='cuda:0'), covar=tensor([0.1038, 0.1545, 0.0534, 0.1789, 0.1287, 0.0246, 0.0659, 0.0569], + device='cuda:0'), in_proj_covar=tensor([0.0300, 0.0324, 0.0289, 0.0318, 0.0316, 0.0275, 0.0432, 0.0306], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 08:11:45,667 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=195367.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:11:50,114 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=195373.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:12:02,400 INFO [train.py:901] (0/4) Epoch 25, batch 1400, loss[loss=0.1963, simple_loss=0.2905, pruned_loss=0.051, over 8439.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2847, pruned_loss=0.05975, over 1615392.49 frames. ], batch size: 27, lr: 3.04e-03, grad_scale: 4.0 +2023-02-07 08:12:03,971 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=195392.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:12:10,478 INFO [optim.py:369] (0/4) 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 08:12:13,196 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=195406.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 08:12:15,756 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=195410.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:12:30,621 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=195433.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:12:36,126 INFO [train.py:901] (0/4) Epoch 25, batch 1450, loss[loss=0.2409, simple_loss=0.3205, pruned_loss=0.08061, over 8231.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2854, pruned_loss=0.06071, over 1612520.33 frames. ], batch size: 48, lr: 3.04e-03, grad_scale: 4.0 +2023-02-07 08:12:40,059 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 +2023-02-07 08:12:44,146 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-07 08:13:10,158 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 25, batch 1500, loss[loss=0.1821, simple_loss=0.2569, pruned_loss=0.0536, over 7442.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2858, pruned_loss=0.06044, over 1613557.38 frames. ], batch size: 17, lr: 3.04e-03, grad_scale: 4.0 +2023-02-07 08:13:18,980 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 +2023-02-07 08:13:19,800 INFO [optim.py:369] (0/4) 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 08:13:33,935 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195521.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 08:13:46,243 INFO [train.py:901] (0/4) Epoch 25, batch 1550, loss[loss=0.1657, simple_loss=0.2531, pruned_loss=0.03917, over 7801.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2848, pruned_loss=0.05998, over 1612022.02 frames. ], batch size: 20, lr: 3.04e-03, grad_scale: 4.0 +2023-02-07 08:13:51,896 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195548.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:14:05,612 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0921, 1.8770, 2.3238, 1.9961, 2.2074, 2.1605, 1.9929, 1.1568], + device='cuda:0'), covar=tensor([0.5405, 0.4680, 0.2034, 0.3326, 0.2366, 0.3090, 0.2029, 0.5009], + device='cuda:0'), in_proj_covar=tensor([0.0950, 0.1005, 0.0823, 0.0972, 0.1014, 0.0914, 0.0762, 0.0838], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 08:14:14,121 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=195580.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:14:21,350 INFO [train.py:901] (0/4) Epoch 25, batch 1600, loss[loss=0.1842, simple_loss=0.2639, pruned_loss=0.0523, over 7934.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2847, pruned_loss=0.05971, over 1613372.01 frames. ], batch size: 20, lr: 3.04e-03, grad_scale: 8.0 +2023-02-07 08:14:22,241 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8604, 2.1537, 1.7877, 2.5920, 1.3125, 1.6231, 1.9871, 2.0738], + device='cuda:0'), covar=tensor([0.0738, 0.0669, 0.0932, 0.0356, 0.1019, 0.1259, 0.0729, 0.0745], + device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0196, 0.0246, 0.0214, 0.0206, 0.0246, 0.0251, 0.0208], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 08:14:29,476 INFO [optim.py:369] (0/4) 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 08:14:32,310 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=195605.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:14:55,970 INFO [train.py:901] (0/4) Epoch 25, batch 1650, loss[loss=0.2113, simple_loss=0.2716, pruned_loss=0.07554, over 7426.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2849, pruned_loss=0.05964, over 1615718.39 frames. ], batch size: 17, lr: 3.04e-03, grad_scale: 8.0 +2023-02-07 08:15:29,751 INFO [train.py:901] (0/4) Epoch 25, batch 1700, loss[loss=0.2079, simple_loss=0.2851, pruned_loss=0.06532, over 7803.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2854, pruned_loss=0.05971, over 1618641.36 frames. ], batch size: 20, lr: 3.04e-03, grad_scale: 8.0 +2023-02-07 08:15:35,724 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.67 vs. limit=5.0 +2023-02-07 08:15:38,031 INFO [optim.py:369] (0/4) 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 08:15:52,243 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 +2023-02-07 08:16:05,355 INFO [train.py:901] (0/4) Epoch 25, batch 1750, loss[loss=0.2048, simple_loss=0.2857, pruned_loss=0.06194, over 8352.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2857, pruned_loss=0.05986, over 1621226.48 frames. ], batch size: 26, lr: 3.04e-03, grad_scale: 8.0 +2023-02-07 08:16:06,413 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1424, 1.8979, 2.4655, 2.0045, 2.5199, 2.2075, 2.0275, 1.3081], + device='cuda:0'), covar=tensor([0.5840, 0.4968, 0.2100, 0.3863, 0.2450, 0.2998, 0.1903, 0.5401], + device='cuda:0'), in_proj_covar=tensor([0.0953, 0.1008, 0.0825, 0.0976, 0.1017, 0.0917, 0.0764, 0.0841], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 08:16:09,124 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=195744.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:16:09,890 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.58 vs. limit=5.0 +2023-02-07 08:16:15,710 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=195754.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:16:26,033 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=195769.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:16:31,535 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=195777.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 08:16:36,280 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9422, 1.8600, 2.8995, 2.1617, 2.5591, 2.0147, 1.7359, 1.4709], + device='cuda:0'), covar=tensor([0.6788, 0.5936, 0.1820, 0.3982, 0.3082, 0.4113, 0.2898, 0.5405], + device='cuda:0'), in_proj_covar=tensor([0.0954, 0.1008, 0.0825, 0.0977, 0.1016, 0.0917, 0.0764, 0.0841], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 08:16:40,101 INFO [train.py:901] (0/4) Epoch 25, batch 1800, loss[loss=0.1683, simple_loss=0.2515, pruned_loss=0.04254, over 7253.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2854, pruned_loss=0.05979, over 1621592.45 frames. ], batch size: 16, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:16:48,985 INFO [optim.py:369] (0/4) 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 08:16:49,208 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=195802.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 08:16:50,558 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=195804.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:16:56,738 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.72 vs. limit=5.0 +2023-02-07 08:17:07,903 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=195829.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:17:08,883 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 +2023-02-07 08:17:15,238 INFO [train.py:901] (0/4) Epoch 25, batch 1850, loss[loss=0.2097, simple_loss=0.2921, pruned_loss=0.06367, over 8562.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2854, pruned_loss=0.05975, over 1618472.81 frames. ], batch size: 49, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:17:36,411 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195869.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:17:50,156 INFO [train.py:901] (0/4) Epoch 25, batch 1900, loss[loss=0.1911, simple_loss=0.2815, pruned_loss=0.05033, over 8466.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2851, pruned_loss=0.05971, over 1619025.82 frames. ], batch size: 27, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:17:58,362 INFO [optim.py:369] (0/4) 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 08:18:20,510 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5383, 1.4781, 1.8424, 1.1906, 1.2011, 1.8536, 0.2032, 1.1526], + device='cuda:0'), covar=tensor([0.1571, 0.1220, 0.0378, 0.0951, 0.2481, 0.0409, 0.2013, 0.1217], + device='cuda:0'), in_proj_covar=tensor([0.0194, 0.0200, 0.0130, 0.0219, 0.0271, 0.0139, 0.0171, 0.0196], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 08:18:24,462 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 25, batch 1950, loss[loss=0.2067, simple_loss=0.2848, pruned_loss=0.06428, over 8129.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2856, pruned_loss=0.06001, over 1617195.32 frames. ], batch size: 22, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:18:37,850 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-07 08:18:57,284 WARNING [train.py:1067] (0/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-07 08:19:00,593 INFO [train.py:901] (0/4) Epoch 25, batch 2000, loss[loss=0.1936, simple_loss=0.2587, pruned_loss=0.06424, over 7563.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2859, pruned_loss=0.06048, over 1612807.69 frames. ], batch size: 18, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:19:07,497 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-196000.pt +2023-02-07 08:19:09,751 INFO [optim.py:369] (0/4) 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 08:19:35,699 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1172, 1.9698, 2.5793, 2.1475, 2.5768, 2.1582, 2.0268, 1.3987], + device='cuda:0'), covar=tensor([0.5881, 0.5010, 0.2044, 0.3725, 0.2475, 0.3206, 0.2074, 0.5538], + device='cuda:0'), in_proj_covar=tensor([0.0949, 0.1001, 0.0819, 0.0971, 0.1010, 0.0911, 0.0759, 0.0835], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 08:19:36,115 INFO [train.py:901] (0/4) Epoch 25, batch 2050, loss[loss=0.2096, simple_loss=0.3003, pruned_loss=0.05947, over 8747.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2853, pruned_loss=0.05982, over 1615701.32 frames. ], batch size: 30, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:20:11,106 INFO [train.py:901] (0/4) Epoch 25, batch 2100, loss[loss=0.1957, simple_loss=0.2852, pruned_loss=0.05307, over 7811.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2857, pruned_loss=0.0603, over 1618256.38 frames. ], batch size: 20, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:20:20,386 INFO [optim.py:369] (0/4) 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 08:20:35,879 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=196125.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:20:46,046 INFO [train.py:901] (0/4) Epoch 25, batch 2150, loss[loss=0.2206, simple_loss=0.305, pruned_loss=0.0681, over 8626.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2861, pruned_loss=0.0606, over 1618016.49 frames. ], batch size: 34, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:20:54,019 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=196150.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:21:22,081 INFO [train.py:901] (0/4) Epoch 25, batch 2200, loss[loss=0.197, simple_loss=0.2853, pruned_loss=0.05439, over 8288.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2872, pruned_loss=0.06092, over 1619368.53 frames. ], batch size: 23, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:21:30,653 INFO [optim.py:369] (0/4) 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 08:21:56,962 INFO [train.py:901] (0/4) Epoch 25, batch 2250, loss[loss=0.2196, simple_loss=0.3108, pruned_loss=0.06417, over 8244.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2852, pruned_loss=0.05989, over 1614367.05 frames. ], batch size: 24, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:22:32,050 INFO [train.py:901] (0/4) Epoch 25, batch 2300, loss[loss=0.1753, simple_loss=0.2546, pruned_loss=0.04804, over 7410.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2853, pruned_loss=0.05967, over 1614456.14 frames. ], batch size: 17, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:22:40,957 INFO [optim.py:369] (0/4) 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 08:23:07,216 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=196339.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:23:07,741 INFO [train.py:901] (0/4) Epoch 25, batch 2350, loss[loss=0.2296, simple_loss=0.3146, pruned_loss=0.07233, over 8315.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2856, pruned_loss=0.05993, over 1613274.23 frames. ], batch size: 25, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:23:34,279 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=196378.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:23:42,080 INFO [train.py:901] (0/4) Epoch 25, batch 2400, loss[loss=0.1672, simple_loss=0.2475, pruned_loss=0.04348, over 7796.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2852, pruned_loss=0.05977, over 1613936.93 frames. ], batch size: 19, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:23:50,280 INFO [optim.py:369] (0/4) 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 08:24:17,356 INFO [train.py:901] (0/4) Epoch 25, batch 2450, loss[loss=0.1707, simple_loss=0.2632, pruned_loss=0.0391, over 8235.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2857, pruned_loss=0.05974, over 1614037.17 frames. ], batch size: 22, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:24:51,934 INFO [train.py:901] (0/4) Epoch 25, batch 2500, loss[loss=0.2036, simple_loss=0.2837, pruned_loss=0.06175, over 8241.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2855, pruned_loss=0.05976, over 1615037.49 frames. ], batch size: 22, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:25:00,802 INFO [optim.py:369] (0/4) 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 08:25:27,002 INFO [train.py:901] (0/4) Epoch 25, batch 2550, loss[loss=0.1802, simple_loss=0.2567, pruned_loss=0.05183, over 7924.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2856, pruned_loss=0.05979, over 1613165.04 frames. ], batch size: 20, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:25:33,106 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4640, 2.7200, 2.2727, 4.0839, 1.5663, 1.9593, 2.4573, 2.9285], + device='cuda:0'), covar=tensor([0.0707, 0.0831, 0.0872, 0.0246, 0.1063, 0.1254, 0.0928, 0.0756], + device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0196, 0.0244, 0.0212, 0.0205, 0.0247, 0.0249, 0.0206], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 08:26:02,151 INFO [train.py:901] (0/4) Epoch 25, batch 2600, loss[loss=0.1822, simple_loss=0.2541, pruned_loss=0.05516, over 7427.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2854, pruned_loss=0.05995, over 1610460.88 frames. ], batch size: 17, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:26:06,407 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=196596.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:26:10,247 INFO [optim.py:369] (0/4) 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 08:26:18,993 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.92 vs. limit=5.0 +2023-02-07 08:26:37,066 INFO [train.py:901] (0/4) Epoch 25, batch 2650, loss[loss=0.1598, simple_loss=0.2462, pruned_loss=0.03667, over 8087.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2849, pruned_loss=0.05963, over 1608695.18 frames. ], batch size: 21, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:26:56,396 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9911, 1.4523, 3.3621, 1.4697, 2.3861, 3.6876, 3.7819, 3.1874], + device='cuda:0'), covar=tensor([0.1216, 0.1972, 0.0337, 0.2274, 0.1139, 0.0252, 0.0576, 0.0520], + device='cuda:0'), in_proj_covar=tensor([0.0302, 0.0325, 0.0291, 0.0320, 0.0318, 0.0276, 0.0435, 0.0308], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 08:26:58,181 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-02-07 08:27:03,507 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6826, 2.3521, 4.0262, 1.5915, 2.9747, 2.2728, 1.8539, 2.8596], + device='cuda:0'), covar=tensor([0.1951, 0.2659, 0.0697, 0.4724, 0.1941, 0.3325, 0.2330, 0.2367], + device='cuda:0'), in_proj_covar=tensor([0.0532, 0.0619, 0.0555, 0.0657, 0.0653, 0.0601, 0.0548, 0.0636], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 08:27:08,208 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=196683.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:27:12,710 INFO [train.py:901] (0/4) Epoch 25, batch 2700, loss[loss=0.3188, simple_loss=0.3594, pruned_loss=0.1391, over 7326.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2853, pruned_loss=0.05978, over 1613540.38 frames. ], batch size: 72, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:27:20,586 INFO [optim.py:369] (0/4) 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 08:27:22,823 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2080, 1.0571, 1.2620, 1.0564, 1.0261, 1.2900, 0.0955, 0.9604], + device='cuda:0'), covar=tensor([0.1480, 0.1320, 0.0482, 0.0681, 0.2448, 0.0549, 0.1929, 0.1183], + device='cuda:0'), in_proj_covar=tensor([0.0195, 0.0201, 0.0131, 0.0220, 0.0272, 0.0140, 0.0171, 0.0197], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 08:27:33,978 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=196722.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:27:40,540 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 +2023-02-07 08:27:46,998 INFO [train.py:901] (0/4) Epoch 25, batch 2750, loss[loss=0.2275, simple_loss=0.3067, pruned_loss=0.07413, over 8544.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2854, pruned_loss=0.05938, over 1613139.88 frames. ], batch size: 34, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:28:22,174 INFO [train.py:901] (0/4) Epoch 25, batch 2800, loss[loss=0.2179, simple_loss=0.3029, pruned_loss=0.06647, over 8463.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2848, pruned_loss=0.05934, over 1609580.88 frames. ], batch size: 25, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:28:27,861 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=196797.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:28:28,540 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=196798.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:28:31,141 INFO [optim.py:369] (0/4) 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 08:28:40,900 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3712, 3.7213, 2.3461, 3.1312, 2.9798, 2.1173, 3.0059, 3.1986], + device='cuda:0'), covar=tensor([0.1559, 0.0305, 0.1108, 0.0672, 0.0697, 0.1456, 0.1029, 0.0947], + device='cuda:0'), in_proj_covar=tensor([0.0355, 0.0234, 0.0337, 0.0309, 0.0299, 0.0342, 0.0347, 0.0321], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 08:28:54,947 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 25, batch 2850, loss[loss=0.2616, simple_loss=0.3393, pruned_loss=0.09195, over 8688.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2842, pruned_loss=0.05931, over 1605377.68 frames. ], batch size: 49, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:29:19,391 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=196872.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:29:32,393 INFO [train.py:901] (0/4) Epoch 25, batch 2900, loss[loss=0.1874, simple_loss=0.2629, pruned_loss=0.05594, over 7526.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2839, pruned_loss=0.05926, over 1604087.66 frames. ], batch size: 18, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:29:39,419 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=196899.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:29:41,311 INFO [optim.py:369] (0/4) 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 08:29:49,198 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.1535, 1.3063, 4.3309, 1.6078, 3.8319, 3.5860, 3.9298, 3.7757], + device='cuda:0'), covar=tensor([0.0600, 0.5032, 0.0632, 0.4348, 0.1265, 0.1050, 0.0600, 0.0790], + device='cuda:0'), in_proj_covar=tensor([0.0651, 0.0655, 0.0716, 0.0645, 0.0726, 0.0618, 0.0623, 0.0696], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 08:30:08,121 INFO [train.py:901] (0/4) Epoch 25, batch 2950, loss[loss=0.1735, simple_loss=0.2653, pruned_loss=0.04088, over 8323.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2831, pruned_loss=0.05891, over 1605203.88 frames. ], batch size: 25, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:30:08,201 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=196940.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:30:08,827 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-07 08:30:16,154 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.9965, 6.1083, 5.2143, 2.7013, 5.4248, 5.7900, 5.5470, 5.5254], + device='cuda:0'), covar=tensor([0.0572, 0.0342, 0.0891, 0.4321, 0.0754, 0.0704, 0.1073, 0.0459], + device='cuda:0'), in_proj_covar=tensor([0.0533, 0.0450, 0.0437, 0.0545, 0.0435, 0.0454, 0.0429, 0.0400], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 08:30:37,361 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7933, 1.7111, 2.5354, 1.4908, 1.3949, 2.4674, 0.3935, 1.4290], + device='cuda:0'), covar=tensor([0.1411, 0.1146, 0.0272, 0.1257, 0.2242, 0.0386, 0.2030, 0.1349], + device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0200, 0.0131, 0.0220, 0.0271, 0.0139, 0.0170, 0.0197], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 08:30:42,478 INFO [train.py:901] (0/4) Epoch 25, batch 3000, loss[loss=0.2209, simple_loss=0.2823, pruned_loss=0.07974, over 7248.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2828, pruned_loss=0.05919, over 1601671.32 frames. ], batch size: 16, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:30:42,478 INFO [train.py:926] (0/4) Computing validation loss +2023-02-07 08:30:55,639 INFO [train.py:935] (0/4) Epoch 25, validation: loss=0.1722, simple_loss=0.2721, pruned_loss=0.03618, over 944034.00 frames. +2023-02-07 08:30:55,640 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-07 08:31:03,960 INFO [optim.py:369] (0/4) 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 08:31:12,606 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6100, 2.5668, 1.5900, 2.3697, 2.2208, 1.3895, 2.0900, 2.2734], + device='cuda:0'), covar=tensor([0.1546, 0.0533, 0.1543, 0.0673, 0.0851, 0.2072, 0.1195, 0.1006], + device='cuda:0'), in_proj_covar=tensor([0.0357, 0.0237, 0.0339, 0.0311, 0.0301, 0.0344, 0.0349, 0.0323], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 08:31:21,841 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 +2023-02-07 08:31:30,698 INFO [train.py:901] (0/4) Epoch 25, batch 3050, loss[loss=0.1618, simple_loss=0.2487, pruned_loss=0.03746, over 7811.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2829, pruned_loss=0.05928, over 1604324.08 frames. ], batch size: 20, lr: 3.03e-03, grad_scale: 8.0 +2023-02-07 08:31:40,512 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=197054.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:31:41,146 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=197055.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:31:54,610 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 +2023-02-07 08:31:57,855 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=197079.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:32:05,201 INFO [train.py:901] (0/4) Epoch 25, batch 3100, loss[loss=0.2193, simple_loss=0.3012, pruned_loss=0.0687, over 8097.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2842, pruned_loss=0.06006, over 1601599.30 frames. ], batch size: 23, lr: 3.02e-03, grad_scale: 8.0 +2023-02-07 08:32:07,466 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 08:32:13,693 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 +2023-02-07 08:32:24,947 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=197118.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:32:40,196 INFO [train.py:901] (0/4) Epoch 25, batch 3150, loss[loss=0.1917, simple_loss=0.2814, pruned_loss=0.05107, over 8523.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.283, pruned_loss=0.05933, over 1603742.76 frames. ], batch size: 31, lr: 3.02e-03, grad_scale: 8.0 +2023-02-07 08:32:40,969 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=197141.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:32:47,802 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-07 08:32:53,744 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3439, 2.7909, 2.3004, 3.9310, 1.5015, 2.0105, 2.3878, 2.9780], + device='cuda:0'), covar=tensor([0.0740, 0.0735, 0.0777, 0.0249, 0.1099, 0.1195, 0.0995, 0.0751], + device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0197, 0.0245, 0.0213, 0.0206, 0.0247, 0.0251, 0.0207], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 08:33:15,297 INFO [train.py:901] (0/4) Epoch 25, batch 3200, loss[loss=0.1971, simple_loss=0.2853, pruned_loss=0.05447, over 8323.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2831, pruned_loss=0.05932, over 1605584.35 frames. ], batch size: 25, lr: 3.02e-03, grad_scale: 8.0 +2023-02-07 08:33:23,539 INFO [optim.py:369] (0/4) 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 08:33:33,166 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=197216.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:33:50,301 INFO [train.py:901] (0/4) Epoch 25, batch 3250, loss[loss=0.2353, simple_loss=0.322, pruned_loss=0.0743, over 8295.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.283, pruned_loss=0.05926, over 1602898.90 frames. ], batch size: 23, lr: 3.02e-03, grad_scale: 8.0 +2023-02-07 08:33:52,457 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=197256.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:34:25,383 INFO [train.py:901] (0/4) Epoch 25, batch 3300, loss[loss=0.2086, simple_loss=0.2958, pruned_loss=0.06074, over 8524.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2835, pruned_loss=0.05922, over 1606634.49 frames. ], batch size: 28, lr: 3.02e-03, grad_scale: 8.0 +2023-02-07 08:34:34,250 INFO [optim.py:369] (0/4) 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 08:34:40,485 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=197311.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:34:41,135 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6384, 1.9251, 1.9925, 1.3508, 2.0912, 1.4831, 0.7344, 1.8708], + device='cuda:0'), covar=tensor([0.0906, 0.0467, 0.0427, 0.0886, 0.0704, 0.1142, 0.1163, 0.0512], + device='cuda:0'), in_proj_covar=tensor([0.0467, 0.0403, 0.0362, 0.0457, 0.0387, 0.0544, 0.0401, 0.0433], + device='cuda:0'), out_proj_covar=tensor([1.2421e-04, 1.0507e-04, 9.4827e-05, 1.1986e-04, 1.0140e-04, 1.5233e-04, + 1.0755e-04, 1.1386e-04], device='cuda:0') +2023-02-07 08:34:54,230 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=197331.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:34:57,740 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=197336.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:35:00,245 INFO [train.py:901] (0/4) Epoch 25, batch 3350, loss[loss=0.2101, simple_loss=0.2868, pruned_loss=0.06669, over 8328.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2839, pruned_loss=0.05966, over 1608012.74 frames. ], batch size: 26, lr: 3.02e-03, grad_scale: 8.0 +2023-02-07 08:35:13,330 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=197358.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:35:36,174 INFO [train.py:901] (0/4) Epoch 25, batch 3400, loss[loss=0.1887, simple_loss=0.2704, pruned_loss=0.05344, over 8244.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2846, pruned_loss=0.05982, over 1612915.85 frames. ], batch size: 22, lr: 3.02e-03, grad_scale: 16.0 +2023-02-07 08:35:44,268 INFO [optim.py:369] (0/4) 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 08:36:06,293 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=197433.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 08:36:11,060 INFO [train.py:901] (0/4) Epoch 25, batch 3450, loss[loss=0.2286, simple_loss=0.3194, pruned_loss=0.0689, over 8325.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.286, pruned_loss=0.06051, over 1615655.29 frames. ], batch size: 26, lr: 3.02e-03, grad_scale: 16.0 +2023-02-07 08:36:46,261 INFO [train.py:901] (0/4) Epoch 25, batch 3500, loss[loss=0.2401, simple_loss=0.3177, pruned_loss=0.08129, over 8411.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2875, pruned_loss=0.06134, over 1616700.04 frames. ], batch size: 49, lr: 3.02e-03, grad_scale: 16.0 +2023-02-07 08:36:48,493 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0042, 1.6155, 1.4529, 1.4957, 1.3405, 1.3189, 1.2555, 1.2768], + device='cuda:0'), covar=tensor([0.1259, 0.0508, 0.1354, 0.0629, 0.0779, 0.1597, 0.1106, 0.0878], + device='cuda:0'), in_proj_covar=tensor([0.0358, 0.0237, 0.0341, 0.0312, 0.0301, 0.0345, 0.0351, 0.0324], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 08:36:54,913 INFO [optim.py:369] (0/4) 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 08:37:02,029 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=197512.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:37:07,274 WARNING [train.py:1067] (0/4) Exclude cut with ID 453-131332-0000-131866_sp0.9 from training. Duration: 25.3333125 +2023-02-07 08:37:17,727 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=197534.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:37:19,766 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=197537.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:37:21,599 INFO [train.py:901] (0/4) Epoch 25, batch 3550, loss[loss=0.2473, simple_loss=0.3239, pruned_loss=0.08531, over 8348.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2876, pruned_loss=0.0611, over 1616506.69 frames. ], batch size: 24, lr: 3.02e-03, grad_scale: 16.0 +2023-02-07 08:37:31,951 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.2608, 1.9027, 5.4324, 2.5087, 4.9021, 4.5606, 4.9998, 4.8840], + device='cuda:0'), covar=tensor([0.0495, 0.4639, 0.0450, 0.3894, 0.1002, 0.0965, 0.0523, 0.0581], + device='cuda:0'), in_proj_covar=tensor([0.0651, 0.0655, 0.0713, 0.0646, 0.0727, 0.0619, 0.0622, 0.0695], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 08:37:37,370 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=197563.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:37:54,484 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=197587.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:37:56,262 INFO [train.py:901] (0/4) Epoch 25, batch 3600, loss[loss=0.1836, simple_loss=0.2682, pruned_loss=0.04952, over 8248.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2869, pruned_loss=0.0607, over 1615627.64 frames. ], batch size: 22, lr: 3.02e-03, grad_scale: 16.0 +2023-02-07 08:38:05,202 INFO [optim.py:369] (0/4) 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 08:38:12,126 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=197612.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:38:13,530 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=197614.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:38:31,211 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=197639.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:38:31,670 INFO [train.py:901] (0/4) Epoch 25, batch 3650, loss[loss=0.235, simple_loss=0.3277, pruned_loss=0.07117, over 8605.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2865, pruned_loss=0.06032, over 1616788.54 frames. ], batch size: 31, lr: 3.02e-03, grad_scale: 16.0 +2023-02-07 08:39:06,718 INFO [train.py:901] (0/4) Epoch 25, batch 3700, loss[loss=0.1895, simple_loss=0.2797, pruned_loss=0.04968, over 8484.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.286, pruned_loss=0.06012, over 1614460.34 frames. ], batch size: 25, lr: 3.02e-03, grad_scale: 16.0 +2023-02-07 08:39:09,552 WARNING [train.py:1067] (0/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-07 08:39:15,746 INFO [optim.py:369] (0/4) 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 08:39:43,097 INFO [train.py:901] (0/4) Epoch 25, batch 3750, loss[loss=0.1819, simple_loss=0.2652, pruned_loss=0.04931, over 7805.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2855, pruned_loss=0.06011, over 1613041.08 frames. ], batch size: 20, lr: 3.02e-03, grad_scale: 16.0 +2023-02-07 08:40:09,408 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=197777.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 08:40:13,697 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 +2023-02-07 08:40:18,201 INFO [train.py:901] (0/4) Epoch 25, batch 3800, loss[loss=0.2362, simple_loss=0.322, pruned_loss=0.07517, over 8504.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2869, pruned_loss=0.06071, over 1615491.75 frames. ], batch size: 28, lr: 3.02e-03, grad_scale: 16.0 +2023-02-07 08:40:26,494 INFO [optim.py:369] (0/4) 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 08:40:53,469 INFO [train.py:901] (0/4) Epoch 25, batch 3850, loss[loss=0.2098, simple_loss=0.2963, pruned_loss=0.06162, over 8133.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2872, pruned_loss=0.0607, over 1617203.40 frames. ], batch size: 22, lr: 3.02e-03, grad_scale: 16.0 +2023-02-07 08:41:12,912 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-07 08:41:19,583 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=197878.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:41:28,518 INFO [train.py:901] (0/4) Epoch 25, batch 3900, loss[loss=0.1843, simple_loss=0.2764, pruned_loss=0.04615, over 8142.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2851, pruned_loss=0.05928, over 1619015.46 frames. ], batch size: 22, lr: 3.02e-03, grad_scale: 16.0 +2023-02-07 08:41:29,948 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=197892.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 08:41:36,382 INFO [optim.py:369] (0/4) 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 08:41:39,816 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=197907.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:41:41,927 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=197910.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:41:50,299 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.50 vs. limit=2.0 +2023-02-07 08:42:02,719 INFO [train.py:901] (0/4) Epoch 25, batch 3950, loss[loss=0.2053, simple_loss=0.2951, pruned_loss=0.05773, over 8486.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2844, pruned_loss=0.05893, over 1617499.13 frames. ], batch size: 28, lr: 3.02e-03, grad_scale: 16.0 +2023-02-07 08:42:26,595 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 +2023-02-07 08:42:37,828 INFO [train.py:901] (0/4) Epoch 25, batch 4000, loss[loss=0.2009, simple_loss=0.2797, pruned_loss=0.06107, over 7912.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2829, pruned_loss=0.05799, over 1615138.86 frames. ], batch size: 20, lr: 3.02e-03, grad_scale: 16.0 +2023-02-07 08:42:40,157 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=197993.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:42:45,677 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-198000.pt +2023-02-07 08:42:47,782 INFO [optim.py:369] (0/4) 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 08:43:00,833 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.9402, 2.0458, 4.1026, 2.2409, 3.7459, 3.4820, 3.7976, 3.7087], + device='cuda:0'), covar=tensor([0.0664, 0.3745, 0.0807, 0.3825, 0.0900, 0.0903, 0.0608, 0.0602], + device='cuda:0'), in_proj_covar=tensor([0.0651, 0.0656, 0.0718, 0.0645, 0.0726, 0.0621, 0.0624, 0.0696], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 08:43:01,554 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=198022.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:43:07,334 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.5447, 1.7831, 1.8726, 1.6828, 0.9834, 1.7082, 2.1819, 1.7910], + device='cuda:0'), covar=tensor([0.0492, 0.1182, 0.1707, 0.1387, 0.0597, 0.1445, 0.0628, 0.0645], + device='cuda:0'), in_proj_covar=tensor([0.0098, 0.0153, 0.0190, 0.0160, 0.0100, 0.0163, 0.0112, 0.0145], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:0') +2023-02-07 08:43:14,007 INFO [train.py:901] (0/4) Epoch 25, batch 4050, loss[loss=0.1994, simple_loss=0.2875, pruned_loss=0.05572, over 8244.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2828, pruned_loss=0.05801, over 1616058.17 frames. ], batch size: 24, lr: 3.02e-03, grad_scale: 16.0 +2023-02-07 08:43:48,800 INFO [train.py:901] (0/4) Epoch 25, batch 4100, loss[loss=0.1709, simple_loss=0.2745, pruned_loss=0.03369, over 8033.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2831, pruned_loss=0.05764, over 1617798.66 frames. ], batch size: 22, lr: 3.02e-03, grad_scale: 16.0 +2023-02-07 08:43:55,121 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) Epoch 25, batch 4150, loss[loss=0.2067, simple_loss=0.2896, pruned_loss=0.06192, over 8132.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2836, pruned_loss=0.05778, over 1617511.25 frames. ], batch size: 22, lr: 3.02e-03, grad_scale: 16.0 +2023-02-07 08:44:29,932 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=198148.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 08:44:47,413 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198173.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 08:44:58,967 INFO [train.py:901] (0/4) Epoch 25, batch 4200, loss[loss=0.1733, simple_loss=0.2618, pruned_loss=0.04234, over 8188.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2841, pruned_loss=0.05809, over 1617246.19 frames. ], batch size: 23, lr: 3.02e-03, grad_scale: 16.0 +2023-02-07 08:45:08,035 INFO [optim.py:369] (0/4) 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 08:45:09,406 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-07 08:45:33,174 WARNING [train.py:1067] (0/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-07 08:45:35,185 INFO [train.py:901] (0/4) Epoch 25, batch 4250, loss[loss=0.1919, simple_loss=0.2646, pruned_loss=0.05965, over 7800.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2851, pruned_loss=0.05849, over 1617452.09 frames. ], batch size: 19, lr: 3.02e-03, grad_scale: 16.0 +2023-02-07 08:45:41,594 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=198249.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:45:44,716 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=198254.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:45:59,326 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198274.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:46:02,103 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=198278.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:46:09,935 INFO [train.py:901] (0/4) Epoch 25, batch 4300, loss[loss=0.163, simple_loss=0.248, pruned_loss=0.03907, over 8085.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2848, pruned_loss=0.05869, over 1613057.62 frames. ], batch size: 21, lr: 3.02e-03, grad_scale: 16.0 +2023-02-07 08:46:18,877 INFO [optim.py:369] (0/4) 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 08:46:19,845 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198303.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:46:45,628 INFO [train.py:901] (0/4) Epoch 25, batch 4350, loss[loss=0.156, simple_loss=0.2284, pruned_loss=0.04185, over 6820.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2832, pruned_loss=0.05761, over 1613359.51 frames. ], batch size: 15, lr: 3.02e-03, grad_scale: 16.0 +2023-02-07 08:47:04,265 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-07 08:47:06,468 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=198369.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:47:21,572 INFO [train.py:901] (0/4) Epoch 25, batch 4400, loss[loss=0.2168, simple_loss=0.3001, pruned_loss=0.0667, over 8335.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2834, pruned_loss=0.05788, over 1612863.49 frames. ], batch size: 25, lr: 3.01e-03, grad_scale: 16.0 +2023-02-07 08:47:29,521 INFO [optim.py:369] (0/4) 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 08:47:45,274 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 25, batch 4450, loss[loss=0.2231, simple_loss=0.2993, pruned_loss=0.0734, over 8185.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2843, pruned_loss=0.05858, over 1613372.12 frames. ], batch size: 23, lr: 3.01e-03, grad_scale: 16.0 +2023-02-07 08:47:58,932 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=198443.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:48:01,856 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1375, 1.9072, 2.3515, 2.0690, 2.2976, 2.1975, 2.0173, 1.1700], + device='cuda:0'), covar=tensor([0.5915, 0.4637, 0.1982, 0.3617, 0.2542, 0.3209, 0.1919, 0.5118], + device='cuda:0'), in_proj_covar=tensor([0.0945, 0.0998, 0.0816, 0.0969, 0.1010, 0.0911, 0.0757, 0.0835], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 08:48:31,963 INFO [train.py:901] (0/4) Epoch 25, batch 4500, loss[loss=0.223, simple_loss=0.2896, pruned_loss=0.07823, over 7923.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2846, pruned_loss=0.05936, over 1611010.17 frames. ], batch size: 20, lr: 3.01e-03, grad_scale: 16.0 +2023-02-07 08:48:40,439 INFO [optim.py:369] (0/4) 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 08:48:40,470 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-07 08:48:51,900 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=198517.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:49:06,134 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8541, 2.3790, 4.0608, 1.7569, 2.9536, 2.3752, 2.0152, 2.9751], + device='cuda:0'), covar=tensor([0.1922, 0.2806, 0.0949, 0.4716, 0.2118, 0.3275, 0.2439, 0.2437], + device='cuda:0'), in_proj_covar=tensor([0.0533, 0.0622, 0.0557, 0.0658, 0.0657, 0.0605, 0.0552, 0.0639], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 08:49:08,830 INFO [train.py:901] (0/4) Epoch 25, batch 4550, loss[loss=0.1921, simple_loss=0.2782, pruned_loss=0.05298, over 8240.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2826, pruned_loss=0.05884, over 1607525.61 frames. ], batch size: 22, lr: 3.01e-03, grad_scale: 16.0 +2023-02-07 08:49:22,064 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=198558.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:49:24,449 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 +2023-02-07 08:49:44,786 INFO [train.py:901] (0/4) Epoch 25, batch 4600, loss[loss=0.1743, simple_loss=0.2596, pruned_loss=0.04455, over 7719.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2844, pruned_loss=0.06005, over 1612040.41 frames. ], batch size: 18, lr: 3.01e-03, grad_scale: 16.0 +2023-02-07 08:49:52,976 INFO [optim.py:369] (0/4) 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 08:50:09,321 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=198625.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:50:14,653 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=198633.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:50:19,157 INFO [train.py:901] (0/4) Epoch 25, batch 4650, loss[loss=0.1844, simple_loss=0.2801, pruned_loss=0.04434, over 8452.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.284, pruned_loss=0.05958, over 1615112.78 frames. ], batch size: 27, lr: 3.01e-03, grad_scale: 16.0 +2023-02-07 08:50:26,829 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198650.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:50:54,514 INFO [train.py:901] (0/4) Epoch 25, batch 4700, loss[loss=0.1589, simple_loss=0.2489, pruned_loss=0.03441, over 7656.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2843, pruned_loss=0.05929, over 1617496.65 frames. ], batch size: 19, lr: 3.01e-03, grad_scale: 16.0 +2023-02-07 08:50:58,177 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2413, 1.0894, 1.2880, 1.0389, 1.0081, 1.3274, 0.0578, 0.9550], + device='cuda:0'), covar=tensor([0.1418, 0.1209, 0.0535, 0.0676, 0.2427, 0.0517, 0.1993, 0.1276], + device='cuda:0'), in_proj_covar=tensor([0.0196, 0.0202, 0.0132, 0.0222, 0.0275, 0.0141, 0.0172, 0.0198], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 08:51:03,372 INFO [optim.py:369] (0/4) 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 08:51:12,343 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.5809, 1.5063, 1.7991, 1.4472, 0.8925, 1.5829, 2.1113, 1.9601], + device='cuda:0'), covar=tensor([0.0513, 0.1396, 0.1711, 0.1574, 0.0681, 0.1619, 0.0685, 0.0623], + device='cuda:0'), in_proj_covar=tensor([0.0098, 0.0153, 0.0189, 0.0160, 0.0100, 0.0163, 0.0112, 0.0145], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:0') +2023-02-07 08:51:17,651 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2714, 2.4730, 1.9960, 2.9905, 1.4129, 1.7868, 2.2625, 2.4367], + device='cuda:0'), covar=tensor([0.0661, 0.0732, 0.0866, 0.0326, 0.1092, 0.1203, 0.0782, 0.0686], + device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0194, 0.0245, 0.0212, 0.0203, 0.0246, 0.0247, 0.0206], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 08:51:29,737 INFO [train.py:901] (0/4) Epoch 25, batch 4750, loss[loss=0.2148, simple_loss=0.288, pruned_loss=0.07077, over 7686.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2848, pruned_loss=0.05949, over 1620149.39 frames. ], batch size: 18, lr: 3.01e-03, grad_scale: 16.0 +2023-02-07 08:51:42,027 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-07 08:51:45,378 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-07 08:52:05,184 INFO [train.py:901] (0/4) Epoch 25, batch 4800, loss[loss=0.2051, simple_loss=0.2929, pruned_loss=0.05865, over 8475.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.283, pruned_loss=0.05826, over 1619228.89 frames. ], batch size: 25, lr: 3.01e-03, grad_scale: 16.0 +2023-02-07 08:52:13,383 INFO [optim.py:369] (0/4) 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 08:52:22,025 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=198814.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:52:36,111 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-07 08:52:39,633 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198839.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:52:40,103 INFO [train.py:901] (0/4) Epoch 25, batch 4850, loss[loss=0.1906, simple_loss=0.2616, pruned_loss=0.05977, over 7796.00 frames. ], tot_loss[loss=0.199, simple_loss=0.282, pruned_loss=0.05797, over 1613092.00 frames. ], batch size: 19, lr: 3.01e-03, grad_scale: 16.0 +2023-02-07 08:52:54,171 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3202, 2.1133, 1.6612, 1.9576, 1.7416, 1.3331, 1.7429, 1.7319], + device='cuda:0'), covar=tensor([0.1352, 0.0460, 0.1329, 0.0538, 0.0820, 0.1735, 0.0927, 0.0938], + device='cuda:0'), in_proj_covar=tensor([0.0355, 0.0234, 0.0338, 0.0310, 0.0299, 0.0342, 0.0344, 0.0317], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 08:52:55,418 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=198861.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:53:01,725 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=198870.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:53:14,818 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6555, 2.6294, 1.8792, 2.3510, 2.0755, 1.5067, 2.0715, 2.1949], + device='cuda:0'), covar=tensor([0.1299, 0.0399, 0.1246, 0.0516, 0.0771, 0.1692, 0.1002, 0.0847], + device='cuda:0'), in_proj_covar=tensor([0.0357, 0.0235, 0.0340, 0.0311, 0.0301, 0.0343, 0.0346, 0.0319], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 08:53:15,878 INFO [train.py:901] (0/4) Epoch 25, batch 4900, loss[loss=0.2034, simple_loss=0.2872, pruned_loss=0.05985, over 8472.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2818, pruned_loss=0.05844, over 1612414.75 frames. ], batch size: 25, lr: 3.01e-03, grad_scale: 16.0 +2023-02-07 08:53:22,929 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8848, 1.8128, 2.4115, 1.5951, 1.4391, 2.4550, 0.3445, 1.5476], + device='cuda:0'), covar=tensor([0.1830, 0.1291, 0.0432, 0.1098, 0.2447, 0.0423, 0.2074, 0.1171], + device='cuda:0'), in_proj_covar=tensor([0.0196, 0.0201, 0.0132, 0.0221, 0.0274, 0.0141, 0.0171, 0.0198], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 08:53:24,147 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=198901.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:53:24,667 INFO [optim.py:369] (0/4) 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 08:53:50,031 INFO [train.py:901] (0/4) Epoch 25, batch 4950, loss[loss=0.2363, simple_loss=0.3177, pruned_loss=0.07742, over 8194.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2823, pruned_loss=0.05815, over 1616520.77 frames. ], batch size: 23, lr: 3.01e-03, grad_scale: 16.0 +2023-02-07 08:53:54,413 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=198945.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:54:01,873 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9192, 1.7491, 2.5551, 1.7344, 1.4456, 2.5413, 0.4601, 1.6237], + device='cuda:0'), covar=tensor([0.1415, 0.1141, 0.0284, 0.1009, 0.2385, 0.0319, 0.2015, 0.1366], + device='cuda:0'), in_proj_covar=tensor([0.0195, 0.0201, 0.0132, 0.0220, 0.0274, 0.0141, 0.0171, 0.0198], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 08:54:15,936 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=198976.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:54:16,526 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=198977.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:54:25,251 INFO [train.py:901] (0/4) Epoch 25, batch 5000, loss[loss=0.2224, simple_loss=0.302, pruned_loss=0.0714, over 8334.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2827, pruned_loss=0.05805, over 1614046.64 frames. ], batch size: 26, lr: 3.01e-03, grad_scale: 16.0 +2023-02-07 08:54:33,925 INFO [optim.py:369] (0/4) 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 08:54:59,850 INFO [train.py:901] (0/4) Epoch 25, batch 5050, loss[loss=0.2057, simple_loss=0.2908, pruned_loss=0.06029, over 8250.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2824, pruned_loss=0.05802, over 1611507.02 frames. ], batch size: 22, lr: 3.01e-03, grad_scale: 16.0 +2023-02-07 08:55:02,109 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1661, 1.6038, 1.7355, 1.5577, 1.1362, 1.5995, 1.9046, 1.6923], + device='cuda:0'), covar=tensor([0.0567, 0.1255, 0.1685, 0.1447, 0.0631, 0.1456, 0.0708, 0.0632], + device='cuda:0'), in_proj_covar=tensor([0.0099, 0.0153, 0.0190, 0.0161, 0.0100, 0.0163, 0.0113, 0.0145], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:0') +2023-02-07 08:55:14,358 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-07 08:55:35,768 INFO [train.py:901] (0/4) Epoch 25, batch 5100, loss[loss=0.2238, simple_loss=0.323, pruned_loss=0.06232, over 8112.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2827, pruned_loss=0.05832, over 1611102.14 frames. ], batch size: 23, lr: 3.01e-03, grad_scale: 16.0 +2023-02-07 08:55:37,391 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199092.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:55:44,128 INFO [optim.py:369] (0/4) 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 08:56:11,855 INFO [train.py:901] (0/4) Epoch 25, batch 5150, loss[loss=0.1791, simple_loss=0.265, pruned_loss=0.04662, over 7796.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2834, pruned_loss=0.05878, over 1616318.81 frames. ], batch size: 20, lr: 3.01e-03, grad_scale: 16.0 +2023-02-07 08:56:47,047 INFO [train.py:901] (0/4) Epoch 25, batch 5200, loss[loss=0.21, simple_loss=0.2972, pruned_loss=0.06144, over 8243.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2847, pruned_loss=0.05925, over 1616188.77 frames. ], batch size: 22, lr: 3.01e-03, grad_scale: 16.0 +2023-02-07 08:56:49,909 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199194.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:56:55,025 INFO [optim.py:369] (0/4) 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 08:57:04,086 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=199214.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:57:05,109 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 +2023-02-07 08:57:12,750 WARNING [train.py:1067] (0/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-07 08:57:17,170 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199232.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:57:22,321 INFO [train.py:901] (0/4) Epoch 25, batch 5250, loss[loss=0.2078, simple_loss=0.307, pruned_loss=0.05431, over 8530.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2853, pruned_loss=0.06005, over 1612965.42 frames. ], batch size: 28, lr: 3.01e-03, grad_scale: 16.0 +2023-02-07 08:57:25,795 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=199245.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:57:29,645 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 +2023-02-07 08:57:34,066 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.2144, 4.1657, 3.7500, 1.9250, 3.7220, 3.8690, 3.7557, 3.7225], + device='cuda:0'), covar=tensor([0.0812, 0.0591, 0.1101, 0.4369, 0.0916, 0.1174, 0.1374, 0.0802], + device='cuda:0'), in_proj_covar=tensor([0.0542, 0.0456, 0.0442, 0.0555, 0.0439, 0.0460, 0.0432, 0.0404], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 08:57:34,818 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199257.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:57:56,824 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=199289.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:57:57,409 INFO [train.py:901] (0/4) Epoch 25, batch 5300, loss[loss=0.2012, simple_loss=0.2931, pruned_loss=0.05466, over 8342.00 frames. ], tot_loss[loss=0.204, simple_loss=0.287, pruned_loss=0.06047, over 1618544.07 frames. ], batch size: 26, lr: 3.01e-03, grad_scale: 16.0 +2023-02-07 08:58:05,706 INFO [optim.py:369] (0/4) 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 08:58:25,198 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199329.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:58:32,735 INFO [train.py:901] (0/4) Epoch 25, batch 5350, loss[loss=0.2529, simple_loss=0.3352, pruned_loss=0.08529, over 8335.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2854, pruned_loss=0.05948, over 1618656.29 frames. ], batch size: 26, lr: 3.01e-03, grad_scale: 16.0 +2023-02-07 08:58:38,526 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199348.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:58:47,315 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 +2023-02-07 08:58:47,609 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199360.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:58:57,338 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199373.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:59:08,720 INFO [train.py:901] (0/4) Epoch 25, batch 5400, loss[loss=0.2067, simple_loss=0.286, pruned_loss=0.06374, over 8314.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2849, pruned_loss=0.05972, over 1614726.73 frames. ], batch size: 25, lr: 3.01e-03, grad_scale: 32.0 +2023-02-07 08:59:18,148 INFO [optim.py:369] (0/4) 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 08:59:18,346 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199404.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:59:28,341 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199418.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 08:59:41,399 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3574, 2.0557, 2.6926, 2.2482, 2.7684, 2.4008, 2.2067, 1.5064], + device='cuda:0'), covar=tensor([0.5746, 0.5312, 0.2059, 0.3875, 0.2454, 0.3133, 0.1867, 0.5626], + device='cuda:0'), in_proj_covar=tensor([0.0949, 0.1000, 0.0817, 0.0971, 0.1012, 0.0915, 0.0759, 0.0836], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 08:59:43,192 INFO [train.py:901] (0/4) Epoch 25, batch 5450, loss[loss=0.184, simple_loss=0.2597, pruned_loss=0.05416, over 7549.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2838, pruned_loss=0.05874, over 1610420.28 frames. ], batch size: 18, lr: 3.01e-03, grad_scale: 8.0 +2023-02-07 09:00:08,077 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-07 09:00:08,202 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199476.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:00:14,523 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.44 vs. limit=5.0 +2023-02-07 09:00:17,966 INFO [train.py:901] (0/4) Epoch 25, batch 5500, loss[loss=0.2303, simple_loss=0.3026, pruned_loss=0.07905, over 8604.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2829, pruned_loss=0.05833, over 1612437.06 frames. ], batch size: 31, lr: 3.01e-03, grad_scale: 8.0 +2023-02-07 09:00:28,255 INFO [optim.py:369] (0/4) 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 09:00:33,794 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199512.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 09:00:40,195 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 +2023-02-07 09:00:52,114 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 25, batch 5550, loss[loss=0.2265, simple_loss=0.3117, pruned_loss=0.07068, over 8529.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2829, pruned_loss=0.05804, over 1606565.85 frames. ], batch size: 31, lr: 3.01e-03, grad_scale: 8.0 +2023-02-07 09:01:02,087 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199553.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 09:01:24,432 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199585.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:01:27,625 INFO [train.py:901] (0/4) Epoch 25, batch 5600, loss[loss=0.2151, simple_loss=0.296, pruned_loss=0.06711, over 8651.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2835, pruned_loss=0.05878, over 1607315.16 frames. ], batch size: 39, lr: 3.01e-03, grad_scale: 8.0 +2023-02-07 09:01:38,052 INFO [optim.py:369] (0/4) 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 09:01:43,145 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199616.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:02:03,340 INFO [train.py:901] (0/4) Epoch 25, batch 5650, loss[loss=0.1999, simple_loss=0.2874, pruned_loss=0.05622, over 8143.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2842, pruned_loss=0.05906, over 1610556.35 frames. ], batch size: 22, lr: 3.01e-03, grad_scale: 8.0 +2023-02-07 09:02:04,208 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199641.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:02:13,519 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199653.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:02:14,004 WARNING [train.py:1067] (0/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-07 09:02:18,307 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199660.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:02:36,563 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199685.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:02:39,923 INFO [train.py:901] (0/4) Epoch 25, batch 5700, loss[loss=0.1663, simple_loss=0.2445, pruned_loss=0.04405, over 7937.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2829, pruned_loss=0.05813, over 1614784.08 frames. ], batch size: 20, lr: 3.01e-03, grad_scale: 8.0 +2023-02-07 09:02:49,769 INFO [optim.py:369] (0/4) 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 09:03:16,064 INFO [train.py:901] (0/4) Epoch 25, batch 5750, loss[loss=0.1876, simple_loss=0.2715, pruned_loss=0.05191, over 7981.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2807, pruned_loss=0.05721, over 1609993.21 frames. ], batch size: 21, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:03:21,576 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-07 09:03:30,821 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=199762.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:03:50,502 INFO [train.py:901] (0/4) Epoch 25, batch 5800, loss[loss=0.2193, simple_loss=0.3096, pruned_loss=0.06457, over 8669.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2817, pruned_loss=0.05779, over 1610938.23 frames. ], batch size: 39, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:04:00,801 INFO [optim.py:369] (0/4) 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 09:04:11,724 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=199820.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:04:26,580 INFO [train.py:901] (0/4) Epoch 25, batch 5850, loss[loss=0.1885, simple_loss=0.2823, pruned_loss=0.04742, over 8176.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2815, pruned_loss=0.05738, over 1614911.74 frames. ], batch size: 23, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:04:37,359 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=199856.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 09:04:51,709 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199877.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:04:52,571 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.89 vs. limit=5.0 +2023-02-07 09:05:01,001 INFO [train.py:901] (0/4) Epoch 25, batch 5900, loss[loss=0.2821, simple_loss=0.3413, pruned_loss=0.1115, over 8470.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2823, pruned_loss=0.05762, over 1617932.45 frames. ], batch size: 28, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:05:05,785 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=199897.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 09:05:10,359 INFO [optim.py:369] (0/4) 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 09:05:13,991 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199909.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:05:26,436 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199927.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:05:31,326 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199934.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:05:32,028 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 25, batch 5950, loss[loss=0.1709, simple_loss=0.2502, pruned_loss=0.04582, over 7285.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2819, pruned_loss=0.05758, over 1612903.65 frames. ], batch size: 16, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:05:58,231 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199971.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 09:06:11,073 INFO [train.py:901] (0/4) Epoch 25, batch 6000, loss[loss=0.1796, simple_loss=0.2681, pruned_loss=0.04553, over 8248.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.281, pruned_loss=0.05744, over 1610203.82 frames. ], batch size: 22, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:06:11,074 INFO [train.py:926] (0/4) Computing validation loss +2023-02-07 09:06:23,700 INFO [train.py:935] (0/4) Epoch 25, validation: loss=0.1725, simple_loss=0.2721, pruned_loss=0.03643, over 944034.00 frames. +2023-02-07 09:06:23,701 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6722MB +2023-02-07 09:06:30,903 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-200000.pt +2023-02-07 09:06:34,584 INFO [optim.py:369] (0/4) 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 09:06:37,407 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.8010, 5.8972, 5.0266, 2.7156, 5.1310, 5.6254, 5.3468, 5.3110], + device='cuda:0'), covar=tensor([0.0510, 0.0342, 0.0885, 0.3815, 0.0825, 0.0797, 0.1017, 0.0578], + device='cuda:0'), in_proj_covar=tensor([0.0536, 0.0451, 0.0437, 0.0548, 0.0435, 0.0453, 0.0430, 0.0400], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 09:06:40,199 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=200012.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 09:06:49,049 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.5759, 2.9293, 2.4721, 4.0893, 1.7453, 2.1336, 2.7306, 3.0467], + device='cuda:0'), covar=tensor([0.0667, 0.0719, 0.0760, 0.0207, 0.1074, 0.1236, 0.0808, 0.0754], + device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0194, 0.0244, 0.0211, 0.0205, 0.0247, 0.0248, 0.0207], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 09:06:59,670 INFO [train.py:901] (0/4) Epoch 25, batch 6050, loss[loss=0.2081, simple_loss=0.2939, pruned_loss=0.06115, over 8476.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2819, pruned_loss=0.05772, over 1612140.71 frames. ], batch size: 25, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:07:35,041 INFO [train.py:901] (0/4) Epoch 25, batch 6100, loss[loss=0.2165, simple_loss=0.3052, pruned_loss=0.0639, over 8483.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2827, pruned_loss=0.05815, over 1612313.12 frames. ], batch size: 25, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:07:45,362 INFO [optim.py:369] (0/4) 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 09:07:54,362 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-07 09:08:05,858 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=200133.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:08:07,192 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([6.0257, 1.7276, 6.2592, 2.3868, 5.6518, 5.2179, 5.7614, 5.6428], + device='cuda:0'), covar=tensor([0.0493, 0.4710, 0.0309, 0.3823, 0.0969, 0.0850, 0.0452, 0.0496], + device='cuda:0'), in_proj_covar=tensor([0.0659, 0.0657, 0.0725, 0.0648, 0.0733, 0.0623, 0.0625, 0.0700], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 09:08:11,247 INFO [train.py:901] (0/4) Epoch 25, batch 6150, loss[loss=0.2245, simple_loss=0.293, pruned_loss=0.07799, over 7429.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2829, pruned_loss=0.05837, over 1611616.78 frames. ], batch size: 17, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:08:23,399 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=200158.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:08:46,124 INFO [train.py:901] (0/4) Epoch 25, batch 6200, loss[loss=0.1847, simple_loss=0.2567, pruned_loss=0.05639, over 7655.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2834, pruned_loss=0.05872, over 1610108.58 frames. ], batch size: 19, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:08:47,048 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=200191.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:08:49,549 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=200195.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 09:08:55,698 INFO [optim.py:369] (0/4) 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 09:09:04,314 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=200216.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:09:12,408 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1151, 1.8935, 2.4928, 2.0899, 2.5382, 2.2271, 2.0411, 1.3070], + device='cuda:0'), covar=tensor([0.5940, 0.4791, 0.2092, 0.3998, 0.2547, 0.3177, 0.2007, 0.5570], + device='cuda:0'), in_proj_covar=tensor([0.0956, 0.1005, 0.0822, 0.0978, 0.1017, 0.0917, 0.0766, 0.0840], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 09:09:13,086 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=200227.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 09:09:21,660 INFO [train.py:901] (0/4) Epoch 25, batch 6250, loss[loss=0.2595, simple_loss=0.3215, pruned_loss=0.09882, over 6668.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2831, pruned_loss=0.05869, over 1608574.46 frames. ], batch size: 71, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:09:29,289 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 +2023-02-07 09:09:29,782 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=200252.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 09:09:41,361 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=200268.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 09:09:43,274 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=200271.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:09:56,026 INFO [train.py:901] (0/4) Epoch 25, batch 6300, loss[loss=0.206, simple_loss=0.2857, pruned_loss=0.06319, over 8083.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2828, pruned_loss=0.05853, over 1606398.49 frames. ], batch size: 21, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:09:58,159 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=200293.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 09:10:00,769 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=200297.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:10:06,125 INFO [optim.py:369] (0/4) 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 09:10:31,232 INFO [train.py:901] (0/4) Epoch 25, batch 6350, loss[loss=0.1804, simple_loss=0.2775, pruned_loss=0.04167, over 8360.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2822, pruned_loss=0.05856, over 1604105.08 frames. ], batch size: 24, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:11:03,766 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 25, batch 6400, loss[loss=0.1905, simple_loss=0.2891, pruned_loss=0.04591, over 8345.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2824, pruned_loss=0.05851, over 1607731.49 frames. ], batch size: 26, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:11:15,859 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 25, batch 6450, loss[loss=0.1723, simple_loss=0.2753, pruned_loss=0.03469, over 8503.00 frames. ], tot_loss[loss=0.2, simple_loss=0.283, pruned_loss=0.0585, over 1609353.86 frames. ], batch size: 26, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:12:00,417 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 +2023-02-07 09:12:16,064 INFO [train.py:901] (0/4) Epoch 25, batch 6500, loss[loss=0.1606, simple_loss=0.2495, pruned_loss=0.03589, over 7671.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2828, pruned_loss=0.05803, over 1611282.77 frames. ], batch size: 19, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:12:26,035 INFO [optim.py:369] (0/4) 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 09:12:49,387 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=200539.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 09:12:49,943 INFO [train.py:901] (0/4) Epoch 25, batch 6550, loss[loss=0.182, simple_loss=0.262, pruned_loss=0.051, over 8077.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2826, pruned_loss=0.05769, over 1611811.27 frames. ], batch size: 21, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:13:09,851 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-07 09:13:26,022 INFO [train.py:901] (0/4) Epoch 25, batch 6600, loss[loss=0.2056, simple_loss=0.2821, pruned_loss=0.06456, over 8138.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2822, pruned_loss=0.05742, over 1616481.51 frames. ], batch size: 22, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:13:27,811 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 +2023-02-07 09:13:30,820 WARNING [train.py:1067] (0/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-07 09:13:35,550 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 25, batch 6650, loss[loss=0.1897, simple_loss=0.2772, pruned_loss=0.05112, over 8471.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2815, pruned_loss=0.05697, over 1614734.63 frames. ], batch size: 29, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:14:01,585 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=200641.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:14:02,419 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=200642.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:14:10,426 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=200654.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 09:14:16,373 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=200663.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:14:19,897 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=200667.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:14:35,347 INFO [train.py:901] (0/4) Epoch 25, batch 6700, loss[loss=0.2235, simple_loss=0.3074, pruned_loss=0.06982, over 7159.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2811, pruned_loss=0.05694, over 1610633.76 frames. ], batch size: 72, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:14:45,632 INFO [optim.py:369] (0/4) 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 09:15:06,977 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9023, 1.5676, 3.4446, 1.5186, 2.5813, 3.8129, 3.8676, 3.3028], + device='cuda:0'), covar=tensor([0.1233, 0.1776, 0.0294, 0.1960, 0.0968, 0.0222, 0.0434, 0.0508], + device='cuda:0'), in_proj_covar=tensor([0.0303, 0.0324, 0.0288, 0.0318, 0.0318, 0.0275, 0.0434, 0.0306], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 09:15:10,910 INFO [train.py:901] (0/4) Epoch 25, batch 6750, loss[loss=0.1823, simple_loss=0.2689, pruned_loss=0.04785, over 7810.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2816, pruned_loss=0.05713, over 1612268.28 frames. ], batch size: 20, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:15:22,767 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=200756.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:15:47,015 INFO [train.py:901] (0/4) Epoch 25, batch 6800, loss[loss=0.2255, simple_loss=0.3082, pruned_loss=0.07145, over 7936.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2813, pruned_loss=0.05709, over 1613190.87 frames. ], batch size: 20, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:15:51,865 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-07 09:15:56,781 INFO [optim.py:369] (0/4) 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 09:16:21,869 INFO [train.py:901] (0/4) Epoch 25, batch 6850, loss[loss=0.152, simple_loss=0.2313, pruned_loss=0.03632, over 7684.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2811, pruned_loss=0.05719, over 1611605.47 frames. ], batch size: 18, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:16:40,400 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1709, 1.5013, 3.5314, 1.6210, 2.4958, 3.9395, 4.0372, 3.3549], + device='cuda:0'), covar=tensor([0.1049, 0.1904, 0.0258, 0.1982, 0.1076, 0.0199, 0.0426, 0.0525], + device='cuda:0'), in_proj_covar=tensor([0.0302, 0.0324, 0.0288, 0.0318, 0.0317, 0.0274, 0.0433, 0.0306], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 09:16:40,956 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-07 09:16:56,592 INFO [train.py:901] (0/4) Epoch 25, batch 6900, loss[loss=0.1696, simple_loss=0.239, pruned_loss=0.05013, over 7536.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2808, pruned_loss=0.05693, over 1614162.00 frames. ], batch size: 18, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:17:03,070 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 +2023-02-07 09:17:06,816 INFO [optim.py:369] (0/4) 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 09:17:11,227 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=200910.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 09:17:28,172 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=200935.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 09:17:31,355 INFO [train.py:901] (0/4) Epoch 25, batch 6950, loss[loss=0.2344, simple_loss=0.3138, pruned_loss=0.07745, over 8081.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.281, pruned_loss=0.05711, over 1607724.95 frames. ], batch size: 21, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:17:34,436 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 +2023-02-07 09:17:36,186 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7580, 1.7085, 2.2935, 1.5072, 1.3668, 2.2292, 0.3930, 1.4485], + device='cuda:0'), covar=tensor([0.1552, 0.1100, 0.0317, 0.0987, 0.2386, 0.0483, 0.2029, 0.1233], + device='cuda:0'), in_proj_covar=tensor([0.0194, 0.0199, 0.0130, 0.0220, 0.0273, 0.0140, 0.0170, 0.0196], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 09:17:50,976 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-07 09:18:07,799 INFO [train.py:901] (0/4) Epoch 25, batch 7000, loss[loss=0.2151, simple_loss=0.3044, pruned_loss=0.0629, over 8450.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2818, pruned_loss=0.05733, over 1612370.17 frames. ], batch size: 29, lr: 3.00e-03, grad_scale: 8.0 +2023-02-07 09:18:17,567 INFO [optim.py:369] (0/4) 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 09:18:19,768 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=201007.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:18:23,173 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201012.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:18:40,485 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201037.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:18:42,298 INFO [train.py:901] (0/4) Epoch 25, batch 7050, loss[loss=0.1831, simple_loss=0.2789, pruned_loss=0.04364, over 8480.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2823, pruned_loss=0.05735, over 1612333.18 frames. ], batch size: 29, lr: 2.99e-03, grad_scale: 8.0 +2023-02-07 09:19:12,437 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 +2023-02-07 09:19:16,824 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=201089.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:19:17,386 INFO [train.py:901] (0/4) Epoch 25, batch 7100, loss[loss=0.2197, simple_loss=0.2958, pruned_loss=0.07183, over 7703.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2817, pruned_loss=0.05723, over 1610461.27 frames. ], batch size: 18, lr: 2.99e-03, grad_scale: 8.0 +2023-02-07 09:19:26,873 INFO [optim.py:369] (0/4) 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 09:19:33,404 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8055, 1.6477, 2.4019, 1.4234, 1.3077, 2.3490, 0.4178, 1.4737], + device='cuda:0'), covar=tensor([0.1637, 0.1245, 0.0319, 0.1382, 0.2490, 0.0381, 0.2153, 0.1319], + device='cuda:0'), in_proj_covar=tensor([0.0195, 0.0199, 0.0130, 0.0219, 0.0273, 0.0140, 0.0169, 0.0197], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 09:19:40,018 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=201122.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:19:52,247 INFO [train.py:901] (0/4) Epoch 25, batch 7150, loss[loss=0.1649, simple_loss=0.2614, pruned_loss=0.03427, over 7969.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2814, pruned_loss=0.057, over 1611222.08 frames. ], batch size: 21, lr: 2.99e-03, grad_scale: 8.0 +2023-02-07 09:20:00,669 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3641, 1.5538, 2.1769, 1.2786, 1.7142, 1.5921, 1.4729, 1.7395], + device='cuda:0'), covar=tensor([0.1484, 0.2005, 0.0756, 0.3585, 0.1527, 0.2410, 0.1799, 0.2015], + device='cuda:0'), in_proj_covar=tensor([0.0535, 0.0626, 0.0559, 0.0661, 0.0661, 0.0605, 0.0553, 0.0643], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 09:20:23,854 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6444, 1.9944, 3.2019, 1.4730, 2.4031, 2.1664, 1.7235, 2.4940], + device='cuda:0'), covar=tensor([0.1912, 0.2931, 0.0825, 0.4909, 0.2037, 0.3141, 0.2540, 0.2273], + device='cuda:0'), in_proj_covar=tensor([0.0536, 0.0627, 0.0560, 0.0662, 0.0662, 0.0606, 0.0554, 0.0644], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 09:20:28,430 INFO [train.py:901] (0/4) Epoch 25, batch 7200, loss[loss=0.1691, simple_loss=0.2553, pruned_loss=0.04145, over 7817.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2827, pruned_loss=0.0577, over 1614782.34 frames. ], batch size: 20, lr: 2.99e-03, grad_scale: 8.0 +2023-02-07 09:20:38,242 INFO [optim.py:369] (0/4) 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 09:20:46,061 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7981, 1.4834, 2.8900, 1.4558, 2.1566, 3.1110, 3.2457, 2.6170], + device='cuda:0'), covar=tensor([0.1079, 0.1638, 0.0344, 0.2011, 0.0915, 0.0272, 0.0601, 0.0566], + device='cuda:0'), in_proj_covar=tensor([0.0300, 0.0322, 0.0288, 0.0317, 0.0316, 0.0273, 0.0432, 0.0305], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 09:20:55,928 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-02-07 09:21:03,482 INFO [train.py:901] (0/4) Epoch 25, batch 7250, loss[loss=0.2051, simple_loss=0.2862, pruned_loss=0.06199, over 8469.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2837, pruned_loss=0.05802, over 1616375.17 frames. ], batch size: 25, lr: 2.99e-03, grad_scale: 8.0 +2023-02-07 09:21:23,522 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.60 vs. limit=2.0 +2023-02-07 09:21:25,201 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=201271.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:21:37,959 INFO [train.py:901] (0/4) Epoch 25, batch 7300, loss[loss=0.1678, simple_loss=0.2477, pruned_loss=0.04395, over 6429.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2839, pruned_loss=0.05845, over 1615938.88 frames. ], batch size: 14, lr: 2.99e-03, grad_scale: 8.0 +2023-02-07 09:21:39,371 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 09:22:13,165 INFO [train.py:901] (0/4) Epoch 25, batch 7350, loss[loss=0.229, simple_loss=0.2969, pruned_loss=0.08053, over 7156.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2843, pruned_loss=0.05864, over 1616993.04 frames. ], batch size: 71, lr: 2.99e-03, grad_scale: 8.0 +2023-02-07 09:22:17,668 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8013, 2.1284, 3.2175, 1.7015, 2.5704, 2.2116, 1.8198, 2.4496], + device='cuda:0'), covar=tensor([0.1818, 0.2530, 0.0891, 0.4374, 0.1835, 0.3085, 0.2401, 0.2256], + device='cuda:0'), in_proj_covar=tensor([0.0535, 0.0627, 0.0559, 0.0663, 0.0661, 0.0607, 0.0553, 0.0643], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 09:22:20,359 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0386, 2.1758, 1.7680, 2.7551, 1.2646, 1.5566, 1.9321, 2.0628], + device='cuda:0'), covar=tensor([0.0713, 0.0692, 0.0924, 0.0350, 0.1166, 0.1411, 0.0920, 0.0838], + device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0195, 0.0244, 0.0211, 0.0206, 0.0247, 0.0249, 0.0207], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 09:22:39,015 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-07 09:22:40,514 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201378.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:22:48,513 INFO [train.py:901] (0/4) Epoch 25, batch 7400, loss[loss=0.1855, simple_loss=0.2672, pruned_loss=0.05194, over 8229.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2838, pruned_loss=0.0586, over 1616993.61 frames. ], batch size: 22, lr: 2.99e-03, grad_scale: 8.0 +2023-02-07 09:22:57,493 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201403.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:22:57,973 INFO [optim.py:369] (0/4) 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 09:22:58,697 WARNING [train.py:1067] (0/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-07 09:23:19,305 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=201433.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:23:22,167 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3942, 1.6045, 1.6747, 1.0240, 1.6676, 1.3780, 0.2653, 1.5451], + device='cuda:0'), covar=tensor([0.0511, 0.0390, 0.0360, 0.0575, 0.0426, 0.0940, 0.0936, 0.0328], + device='cuda:0'), in_proj_covar=tensor([0.0465, 0.0404, 0.0363, 0.0461, 0.0392, 0.0547, 0.0403, 0.0433], + device='cuda:0'), out_proj_covar=tensor([1.2330e-04, 1.0519e-04, 9.4840e-05, 1.2081e-04, 1.0278e-04, 1.5312e-04, + 1.0792e-04, 1.1392e-04], device='cuda:0') +2023-02-07 09:23:23,993 INFO [train.py:901] (0/4) Epoch 25, batch 7450, loss[loss=0.2295, simple_loss=0.3215, pruned_loss=0.06871, over 8343.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.284, pruned_loss=0.05889, over 1614896.64 frames. ], batch size: 26, lr: 2.99e-03, grad_scale: 16.0 +2023-02-07 09:23:38,548 WARNING [train.py:1067] (0/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-07 09:23:59,281 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9548, 2.0748, 1.8765, 2.6094, 1.2368, 1.6575, 1.9049, 1.9993], + device='cuda:0'), covar=tensor([0.0740, 0.0733, 0.0829, 0.0397, 0.1027, 0.1260, 0.0785, 0.0753], + device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0196, 0.0245, 0.0212, 0.0206, 0.0248, 0.0250, 0.0206], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 09:23:59,798 INFO [train.py:901] (0/4) Epoch 25, batch 7500, loss[loss=0.1714, simple_loss=0.2622, pruned_loss=0.04037, over 7922.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2832, pruned_loss=0.05836, over 1614614.56 frames. ], batch size: 20, lr: 2.99e-03, grad_scale: 8.0 +2023-02-07 09:24:09,389 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6477, 1.6013, 2.0712, 1.3511, 1.2786, 2.0609, 0.4496, 1.4268], + device='cuda:0'), covar=tensor([0.1475, 0.1001, 0.0321, 0.0812, 0.2075, 0.0361, 0.1606, 0.1088], + device='cuda:0'), in_proj_covar=tensor([0.0196, 0.0200, 0.0130, 0.0220, 0.0274, 0.0141, 0.0171, 0.0198], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 09:24:09,794 INFO [optim.py:369] (0/4) 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 09:24:25,714 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 +2023-02-07 09:24:34,711 INFO [train.py:901] (0/4) Epoch 25, batch 7550, loss[loss=0.2006, simple_loss=0.2977, pruned_loss=0.05173, over 8640.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2834, pruned_loss=0.05832, over 1618580.52 frames. ], batch size: 34, lr: 2.99e-03, grad_scale: 8.0 +2023-02-07 09:24:40,276 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=201548.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:25:09,417 INFO [train.py:901] (0/4) Epoch 25, batch 7600, loss[loss=0.1959, simple_loss=0.2838, pruned_loss=0.05401, over 8448.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2832, pruned_loss=0.05799, over 1621130.22 frames. ], batch size: 25, lr: 2.99e-03, grad_scale: 8.0 +2023-02-07 09:25:20,540 INFO [optim.py:369] (0/4) 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 09:25:27,310 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=201615.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:25:41,303 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 25, batch 7650, loss[loss=0.2166, simple_loss=0.3015, pruned_loss=0.06588, over 8348.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2844, pruned_loss=0.05909, over 1620801.27 frames. ], batch size: 26, lr: 2.99e-03, grad_scale: 8.0 +2023-02-07 09:26:00,397 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=201662.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:26:12,134 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=201679.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:26:14,193 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0422, 2.5310, 2.7078, 1.4827, 2.8663, 1.5843, 1.5374, 2.0401], + device='cuda:0'), covar=tensor([0.1043, 0.0454, 0.0449, 0.0995, 0.0593, 0.1151, 0.1247, 0.0703], + device='cuda:0'), in_proj_covar=tensor([0.0465, 0.0404, 0.0362, 0.0460, 0.0393, 0.0545, 0.0404, 0.0433], + device='cuda:0'), out_proj_covar=tensor([1.2350e-04, 1.0521e-04, 9.4487e-05, 1.2071e-04, 1.0281e-04, 1.5270e-04, + 1.0792e-04, 1.1377e-04], device='cuda:0') +2023-02-07 09:26:19,453 INFO [train.py:901] (0/4) Epoch 25, batch 7700, loss[loss=0.1932, simple_loss=0.2695, pruned_loss=0.05846, over 7418.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.283, pruned_loss=0.05833, over 1618346.25 frames. ], batch size: 17, lr: 2.99e-03, grad_scale: 8.0 +2023-02-07 09:26:30,382 INFO [optim.py:369] (0/4) 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 09:26:43,967 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=201724.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:26:48,005 INFO [zipformer.py:1185] (0/4) 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] (0/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-07 09:26:54,627 INFO [train.py:901] (0/4) Epoch 25, batch 7750, loss[loss=0.2036, simple_loss=0.288, pruned_loss=0.05959, over 8365.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2826, pruned_loss=0.05811, over 1616044.25 frames. ], batch size: 24, lr: 2.99e-03, grad_scale: 8.0 +2023-02-07 09:27:02,275 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=201751.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:27:20,009 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.8312, 2.4349, 3.7877, 1.9090, 1.9385, 3.7796, 0.5017, 2.1240], + device='cuda:0'), covar=tensor([0.1586, 0.1108, 0.0170, 0.1622, 0.2386, 0.0214, 0.2322, 0.1368], + device='cuda:0'), in_proj_covar=tensor([0.0197, 0.0201, 0.0131, 0.0221, 0.0275, 0.0141, 0.0171, 0.0199], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 09:27:20,636 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.8180, 5.9344, 5.1021, 2.6040, 5.2255, 5.6023, 5.3173, 5.4168], + device='cuda:0'), covar=tensor([0.0587, 0.0382, 0.0944, 0.4434, 0.0756, 0.0934, 0.1216, 0.0612], + device='cuda:0'), in_proj_covar=tensor([0.0533, 0.0452, 0.0439, 0.0549, 0.0436, 0.0457, 0.0430, 0.0397], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 09:27:29,977 INFO [train.py:901] (0/4) Epoch 25, batch 7800, loss[loss=0.2026, simple_loss=0.293, pruned_loss=0.05608, over 8791.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2826, pruned_loss=0.05756, over 1619985.87 frames. ], batch size: 40, lr: 2.99e-03, grad_scale: 8.0 +2023-02-07 09:27:40,035 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201804.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:27:40,476 INFO [optim.py:369] (0/4) 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 09:27:56,321 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201829.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:28:03,384 INFO [train.py:901] (0/4) Epoch 25, batch 7850, loss[loss=0.2171, simple_loss=0.3022, pruned_loss=0.06601, over 8607.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2833, pruned_loss=0.05773, over 1620096.98 frames. ], batch size: 31, lr: 2.99e-03, grad_scale: 4.0 +2023-02-07 09:28:26,017 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 +2023-02-07 09:28:36,517 INFO [train.py:901] (0/4) Epoch 25, batch 7900, loss[loss=0.1848, simple_loss=0.2621, pruned_loss=0.05377, over 7774.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2835, pruned_loss=0.05826, over 1614699.44 frames. ], batch size: 19, lr: 2.99e-03, grad_scale: 4.0 +2023-02-07 09:28:47,141 INFO [optim.py:369] (0/4) 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 09:29:09,581 INFO [train.py:901] (0/4) Epoch 25, batch 7950, loss[loss=0.1652, simple_loss=0.2528, pruned_loss=0.03879, over 7257.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2839, pruned_loss=0.05838, over 1615748.05 frames. ], batch size: 16, lr: 2.99e-03, grad_scale: 4.0 +2023-02-07 09:29:40,304 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201986.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:29:42,800 INFO [train.py:901] (0/4) Epoch 25, batch 8000, loss[loss=0.1762, simple_loss=0.253, pruned_loss=0.04964, over 7426.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2844, pruned_loss=0.05907, over 1613912.42 frames. ], batch size: 17, lr: 2.99e-03, grad_scale: 8.0 +2023-02-07 09:29:49,508 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-202000.pt +2023-02-07 09:29:54,397 INFO [optim.py:369] (0/4) 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 09:29:54,493 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=202006.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:29:55,326 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=202007.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:29:55,934 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=202008.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:29:58,060 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=202011.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:30:06,541 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=202023.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:30:07,922 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.8233, 1.7684, 2.0055, 1.8317, 1.2167, 1.7668, 2.3891, 2.1649], + device='cuda:0'), covar=tensor([0.0439, 0.1202, 0.1534, 0.1297, 0.0558, 0.1374, 0.0582, 0.0571], + device='cuda:0'), in_proj_covar=tensor([0.0098, 0.0153, 0.0189, 0.0160, 0.0100, 0.0163, 0.0112, 0.0145], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:0') +2023-02-07 09:30:12,410 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=202032.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:30:16,032 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 +2023-02-07 09:30:17,710 INFO [train.py:901] (0/4) Epoch 25, batch 8050, loss[loss=0.2306, simple_loss=0.3037, pruned_loss=0.07876, over 6944.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2842, pruned_loss=0.05992, over 1601560.21 frames. ], batch size: 72, lr: 2.99e-03, grad_scale: 8.0 +2023-02-07 09:30:34,607 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-02-07 09:30:36,895 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=202068.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:30:40,703 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/epoch-25.pt +2023-02-07 09:30:51,763 WARNING [train.py:1067] (0/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-07 09:30:55,055 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 +2023-02-07 09:30:55,313 INFO [train.py:901] (0/4) Epoch 26, batch 0, loss[loss=0.2188, simple_loss=0.2964, pruned_loss=0.07057, over 8255.00 frames. ], tot_loss[loss=0.2188, simple_loss=0.2964, pruned_loss=0.07057, over 8255.00 frames. ], batch size: 24, lr: 2.93e-03, grad_scale: 8.0 +2023-02-07 09:30:55,314 INFO [train.py:926] (0/4) Computing validation loss +2023-02-07 09:31:04,569 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2914, 2.0870, 1.6358, 1.8753, 1.7637, 1.5211, 1.6988, 1.6868], + device='cuda:0'), covar=tensor([0.1547, 0.0450, 0.1327, 0.0618, 0.0707, 0.1601, 0.1044, 0.1077], + device='cuda:0'), in_proj_covar=tensor([0.0358, 0.0239, 0.0341, 0.0314, 0.0303, 0.0346, 0.0350, 0.0322], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 09:31:06,904 INFO [train.py:935] (0/4) 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] (0/4) Maximum memory allocated so far is 6722MB +2023-02-07 09:31:21,600 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-07 09:31:29,813 INFO [optim.py:369] (0/4) 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 09:31:36,234 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9699, 2.1554, 1.8091, 2.7895, 1.3164, 1.6549, 2.0754, 2.1721], + device='cuda:0'), covar=tensor([0.0746, 0.0786, 0.0948, 0.0358, 0.1061, 0.1303, 0.0784, 0.0755], + device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0195, 0.0245, 0.0211, 0.0204, 0.0245, 0.0248, 0.0205], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 09:31:40,828 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=202121.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:31:41,325 INFO [train.py:901] (0/4) Epoch 26, batch 50, loss[loss=0.2382, simple_loss=0.3126, pruned_loss=0.08193, over 8364.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2847, pruned_loss=0.06069, over 364506.77 frames. ], batch size: 26, lr: 2.93e-03, grad_scale: 8.0 +2023-02-07 09:31:52,552 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=202138.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:31:55,750 WARNING [train.py:1067] (0/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-07 09:32:14,786 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5239, 1.4841, 1.8254, 1.2396, 1.1941, 1.8398, 0.1925, 1.2606], + device='cuda:0'), covar=tensor([0.1307, 0.1179, 0.0391, 0.0824, 0.2197, 0.0404, 0.1747, 0.1051], + device='cuda:0'), in_proj_covar=tensor([0.0197, 0.0202, 0.0131, 0.0223, 0.0276, 0.0142, 0.0172, 0.0199], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 09:32:15,972 INFO [train.py:901] (0/4) Epoch 26, batch 100, loss[loss=0.1662, simple_loss=0.2526, pruned_loss=0.03992, over 7810.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2854, pruned_loss=0.05918, over 643781.11 frames. ], batch size: 19, lr: 2.93e-03, grad_scale: 8.0 +2023-02-07 09:32:18,604 WARNING [train.py:1067] (0/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-07 09:32:23,611 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=202183.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:32:40,584 INFO [optim.py:369] (0/4) 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 09:32:51,107 INFO [train.py:901] (0/4) Epoch 26, batch 150, loss[loss=0.1923, simple_loss=0.284, pruned_loss=0.05028, over 8572.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2853, pruned_loss=0.06027, over 859182.62 frames. ], batch size: 34, lr: 2.93e-03, grad_scale: 8.0 +2023-02-07 09:33:22,870 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 +2023-02-07 09:33:26,388 INFO [train.py:901] (0/4) Epoch 26, batch 200, loss[loss=0.1969, simple_loss=0.2871, pruned_loss=0.05337, over 8556.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2853, pruned_loss=0.0601, over 1026563.52 frames. ], batch size: 39, lr: 2.93e-03, grad_scale: 8.0 +2023-02-07 09:33:49,944 INFO [optim.py:369] (0/4) 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 09:34:01,568 INFO [train.py:901] (0/4) Epoch 26, batch 250, loss[loss=0.2352, simple_loss=0.3112, pruned_loss=0.07959, over 8479.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2848, pruned_loss=0.06026, over 1154902.38 frames. ], batch size: 49, lr: 2.93e-03, grad_scale: 8.0 +2023-02-07 09:34:05,778 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.7328, 5.8697, 5.0604, 2.5666, 5.1290, 5.5028, 5.3355, 5.3767], + device='cuda:0'), covar=tensor([0.0484, 0.0386, 0.0938, 0.4159, 0.0710, 0.0819, 0.0933, 0.0541], + device='cuda:0'), in_proj_covar=tensor([0.0534, 0.0451, 0.0437, 0.0549, 0.0435, 0.0455, 0.0427, 0.0397], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 09:34:09,732 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-07 09:34:19,969 WARNING [train.py:1067] (0/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-07 09:34:22,758 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=202352.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:34:36,408 INFO [train.py:901] (0/4) Epoch 26, batch 300, loss[loss=0.1967, simple_loss=0.2782, pruned_loss=0.05762, over 7649.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2857, pruned_loss=0.05995, over 1260275.29 frames. ], batch size: 19, lr: 2.93e-03, grad_scale: 8.0 +2023-02-07 09:34:38,648 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1279, 3.6206, 2.2620, 2.8360, 2.6651, 2.0228, 2.7695, 2.9952], + device='cuda:0'), covar=tensor([0.1763, 0.0368, 0.1257, 0.0760, 0.0886, 0.1688, 0.1088, 0.1150], + device='cuda:0'), in_proj_covar=tensor([0.0360, 0.0239, 0.0344, 0.0315, 0.0304, 0.0349, 0.0351, 0.0324], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 09:34:40,010 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=202377.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:34:52,230 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=202394.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:34:57,255 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=202402.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:34:59,615 INFO [optim.py:369] (0/4) 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 09:35:08,713 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=202419.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:35:10,557 INFO [train.py:901] (0/4) Epoch 26, batch 350, loss[loss=0.1641, simple_loss=0.2437, pruned_loss=0.04228, over 7799.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2839, pruned_loss=0.05882, over 1340236.99 frames. ], batch size: 19, lr: 2.93e-03, grad_scale: 4.0 +2023-02-07 09:35:23,417 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=202439.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:35:40,989 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=202464.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:35:43,088 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=202467.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:35:46,422 INFO [train.py:901] (0/4) Epoch 26, batch 400, loss[loss=0.2108, simple_loss=0.2817, pruned_loss=0.06997, over 7919.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2839, pruned_loss=0.05892, over 1404415.95 frames. ], batch size: 20, lr: 2.93e-03, grad_scale: 8.0 +2023-02-07 09:36:11,056 INFO [optim.py:369] (0/4) 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 09:36:11,213 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.9374, 1.4119, 6.0150, 2.1521, 5.4251, 4.9596, 5.5745, 5.4573], + device='cuda:0'), covar=tensor([0.0427, 0.5631, 0.0391, 0.4077, 0.1032, 0.0954, 0.0469, 0.0514], + device='cuda:0'), in_proj_covar=tensor([0.0658, 0.0654, 0.0720, 0.0645, 0.0731, 0.0626, 0.0626, 0.0697], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 09:36:21,078 INFO [train.py:901] (0/4) Epoch 26, batch 450, loss[loss=0.2249, simple_loss=0.3188, pruned_loss=0.06556, over 8499.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2835, pruned_loss=0.05885, over 1448898.77 frames. ], batch size: 28, lr: 2.92e-03, grad_scale: 8.0 +2023-02-07 09:36:55,495 INFO [train.py:901] (0/4) Epoch 26, batch 500, loss[loss=0.1935, simple_loss=0.2762, pruned_loss=0.0554, over 8568.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2837, pruned_loss=0.05863, over 1488019.06 frames. ], batch size: 31, lr: 2.92e-03, grad_scale: 8.0 +2023-02-07 09:37:11,980 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3879, 1.5770, 1.6403, 1.1289, 1.6439, 1.3300, 0.3178, 1.6231], + device='cuda:0'), covar=tensor([0.0506, 0.0394, 0.0298, 0.0525, 0.0445, 0.0794, 0.0858, 0.0262], + device='cuda:0'), in_proj_covar=tensor([0.0463, 0.0402, 0.0357, 0.0455, 0.0387, 0.0540, 0.0398, 0.0429], + device='cuda:0'), out_proj_covar=tensor([1.2293e-04, 1.0473e-04, 9.3178e-05, 1.1921e-04, 1.0137e-04, 1.5109e-04, + 1.0650e-04, 1.1288e-04], device='cuda:0') +2023-02-07 09:37:19,245 INFO [optim.py:369] (0/4) 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 09:37:19,505 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8445, 1.8848, 3.2716, 2.3332, 2.8407, 1.9562, 1.6789, 1.6164], + device='cuda:0'), covar=tensor([0.7916, 0.6969, 0.2155, 0.4572, 0.3349, 0.4753, 0.3148, 0.6510], + device='cuda:0'), in_proj_covar=tensor([0.0950, 0.1005, 0.0820, 0.0977, 0.1014, 0.0915, 0.0762, 0.0838], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 09:37:29,369 INFO [train.py:901] (0/4) Epoch 26, batch 550, loss[loss=0.2202, simple_loss=0.3084, pruned_loss=0.06603, over 8493.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2849, pruned_loss=0.05896, over 1522337.04 frames. ], batch size: 26, lr: 2.92e-03, grad_scale: 8.0 +2023-02-07 09:38:05,142 INFO [train.py:901] (0/4) Epoch 26, batch 600, loss[loss=0.2281, simple_loss=0.3104, pruned_loss=0.07296, over 8289.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2849, pruned_loss=0.05916, over 1546139.62 frames. ], batch size: 23, lr: 2.92e-03, grad_scale: 8.0 +2023-02-07 09:38:21,738 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-07 09:38:27,323 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1957, 2.5570, 2.7963, 1.6375, 3.0344, 1.8477, 1.5958, 2.1278], + device='cuda:0'), covar=tensor([0.0946, 0.0416, 0.0296, 0.0889, 0.0483, 0.0794, 0.0952, 0.0626], + device='cuda:0'), in_proj_covar=tensor([0.0462, 0.0400, 0.0355, 0.0453, 0.0386, 0.0537, 0.0396, 0.0427], + device='cuda:0'), out_proj_covar=tensor([1.2263e-04, 1.0416e-04, 9.2798e-05, 1.1873e-04, 1.0103e-04, 1.5032e-04, + 1.0597e-04, 1.1237e-04], device='cuda:0') +2023-02-07 09:38:29,023 INFO [optim.py:369] (0/4) 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 09:38:38,958 INFO [train.py:901] (0/4) Epoch 26, batch 650, loss[loss=0.1746, simple_loss=0.2496, pruned_loss=0.0498, over 7522.00 frames. ], tot_loss[loss=0.201, simple_loss=0.284, pruned_loss=0.05899, over 1558270.59 frames. ], batch size: 18, lr: 2.92e-03, grad_scale: 8.0 +2023-02-07 09:38:39,861 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=202723.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:38:57,380 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=202748.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:39:14,890 INFO [train.py:901] (0/4) Epoch 26, batch 700, loss[loss=0.1719, simple_loss=0.247, pruned_loss=0.04843, over 7799.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2847, pruned_loss=0.05971, over 1572505.21 frames. ], batch size: 19, lr: 2.92e-03, grad_scale: 8.0 +2023-02-07 09:39:38,607 INFO [optim.py:369] (0/4) 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 09:39:43,670 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2567, 1.5307, 4.5842, 1.7997, 2.5107, 5.1904, 5.2247, 4.4124], + device='cuda:0'), covar=tensor([0.1220, 0.1916, 0.0272, 0.2093, 0.1214, 0.0160, 0.0475, 0.0592], + device='cuda:0'), in_proj_covar=tensor([0.0301, 0.0325, 0.0287, 0.0318, 0.0317, 0.0273, 0.0433, 0.0305], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 09:39:49,890 INFO [train.py:901] (0/4) Epoch 26, batch 750, loss[loss=0.2691, simple_loss=0.3311, pruned_loss=0.1035, over 6794.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2853, pruned_loss=0.05962, over 1587922.29 frames. ], batch size: 71, lr: 2.92e-03, grad_scale: 8.0 +2023-02-07 09:40:01,300 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 +2023-02-07 09:40:05,054 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-07 09:40:07,991 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=202848.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 09:40:13,831 WARNING [train.py:1067] (0/4) Exclude cut with ID 8631-249866-0030-64025_sp0.9 from training. Duration: 26.32775 +2023-02-07 09:40:24,779 INFO [train.py:901] (0/4) Epoch 26, batch 800, loss[loss=0.24, simple_loss=0.3171, pruned_loss=0.08146, over 8456.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.285, pruned_loss=0.05968, over 1591402.01 frames. ], batch size: 27, lr: 2.92e-03, grad_scale: 8.0 +2023-02-07 09:40:49,974 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 26, batch 850, loss[loss=0.2218, simple_loss=0.3047, pruned_loss=0.06945, over 8284.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2856, pruned_loss=0.05956, over 1596247.50 frames. ], batch size: 23, lr: 2.92e-03, grad_scale: 8.0 +2023-02-07 09:41:33,349 INFO [train.py:901] (0/4) Epoch 26, batch 900, loss[loss=0.1888, simple_loss=0.2832, pruned_loss=0.04715, over 8327.00 frames. ], tot_loss[loss=0.204, simple_loss=0.287, pruned_loss=0.06053, over 1598148.97 frames. ], batch size: 25, lr: 2.92e-03, grad_scale: 8.0 +2023-02-07 09:41:58,978 INFO [optim.py:369] (0/4) 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 09:42:08,859 INFO [train.py:901] (0/4) Epoch 26, batch 950, loss[loss=0.1758, simple_loss=0.2611, pruned_loss=0.04526, over 8320.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2865, pruned_loss=0.06011, over 1606327.23 frames. ], batch size: 25, lr: 2.92e-03, grad_scale: 8.0 +2023-02-07 09:42:31,716 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-07 09:42:32,482 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 26, batch 1000, loss[loss=0.1983, simple_loss=0.2892, pruned_loss=0.05372, over 8660.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2848, pruned_loss=0.05935, over 1606087.41 frames. ], batch size: 34, lr: 2.92e-03, grad_scale: 8.0 +2023-02-07 09:43:04,689 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=203103.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:43:05,260 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-07 09:43:07,153 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 26, batch 1050, loss[loss=0.1817, simple_loss=0.2435, pruned_loss=0.05994, over 7710.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2848, pruned_loss=0.05981, over 1607915.80 frames. ], batch size: 18, lr: 2.92e-03, grad_scale: 8.0 +2023-02-07 09:43:18,682 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-07 09:43:53,212 INFO [train.py:901] (0/4) Epoch 26, batch 1100, loss[loss=0.1757, simple_loss=0.2646, pruned_loss=0.04336, over 7815.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2843, pruned_loss=0.05899, over 1614300.53 frames. ], batch size: 19, lr: 2.92e-03, grad_scale: 8.0 +2023-02-07 09:43:55,400 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.8192, 6.0221, 5.1536, 2.3418, 5.2420, 5.5883, 5.4467, 5.4560], + device='cuda:0'), covar=tensor([0.0505, 0.0328, 0.0971, 0.4523, 0.0762, 0.0723, 0.0978, 0.0466], + device='cuda:0'), in_proj_covar=tensor([0.0530, 0.0451, 0.0438, 0.0549, 0.0435, 0.0453, 0.0428, 0.0395], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 09:44:06,850 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=203192.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 09:44:16,728 INFO [optim.py:369] (0/4) 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 09:44:21,618 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.9238, 2.5198, 3.6951, 1.8778, 1.9430, 3.6163, 0.7774, 2.1808], + device='cuda:0'), covar=tensor([0.1434, 0.1075, 0.0231, 0.1751, 0.2394, 0.0330, 0.1971, 0.1393], + device='cuda:0'), in_proj_covar=tensor([0.0197, 0.0203, 0.0132, 0.0222, 0.0277, 0.0143, 0.0171, 0.0199], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 09:44:27,590 INFO [train.py:901] (0/4) Epoch 26, batch 1150, loss[loss=0.2317, simple_loss=0.3174, pruned_loss=0.07296, over 8281.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2833, pruned_loss=0.05851, over 1616238.95 frames. ], batch size: 23, lr: 2.92e-03, grad_scale: 8.0 +2023-02-07 09:44:27,602 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-07 09:45:02,742 INFO [train.py:901] (0/4) Epoch 26, batch 1200, loss[loss=0.2056, simple_loss=0.2811, pruned_loss=0.0651, over 8083.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2841, pruned_loss=0.05857, over 1613412.45 frames. ], batch size: 21, lr: 2.92e-03, grad_scale: 8.0 +2023-02-07 09:45:27,263 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=203307.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 09:45:27,742 INFO [optim.py:369] (0/4) 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 09:45:36,616 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7094, 1.9096, 2.7611, 1.5799, 1.9902, 2.1077, 1.7225, 1.9749], + device='cuda:0'), covar=tensor([0.1838, 0.2834, 0.0898, 0.4874, 0.2075, 0.3260, 0.2494, 0.2285], + device='cuda:0'), in_proj_covar=tensor([0.0535, 0.0625, 0.0556, 0.0660, 0.0658, 0.0606, 0.0555, 0.0642], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 09:45:37,088 INFO [train.py:901] (0/4) Epoch 26, batch 1250, loss[loss=0.1693, simple_loss=0.2426, pruned_loss=0.04801, over 7526.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2839, pruned_loss=0.0586, over 1612492.95 frames. ], batch size: 18, lr: 2.92e-03, grad_scale: 4.0 +2023-02-07 09:45:37,263 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8853, 1.3836, 1.5626, 1.3491, 1.0507, 1.3986, 1.5827, 1.2711], + device='cuda:0'), covar=tensor([0.0551, 0.1340, 0.1750, 0.1518, 0.0618, 0.1516, 0.0756, 0.0735], + device='cuda:0'), in_proj_covar=tensor([0.0099, 0.0153, 0.0190, 0.0161, 0.0099, 0.0163, 0.0112, 0.0145], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:0') +2023-02-07 09:45:49,531 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8893, 2.1448, 2.2509, 1.3944, 2.3442, 1.6962, 0.7635, 2.0412], + device='cuda:0'), covar=tensor([0.0605, 0.0356, 0.0309, 0.0690, 0.0443, 0.0904, 0.0970, 0.0360], + device='cuda:0'), in_proj_covar=tensor([0.0461, 0.0400, 0.0356, 0.0453, 0.0387, 0.0539, 0.0395, 0.0428], + device='cuda:0'), out_proj_covar=tensor([1.2258e-04, 1.0409e-04, 9.2922e-05, 1.1871e-04, 1.0138e-04, 1.5074e-04, + 1.0581e-04, 1.1255e-04], device='cuda:0') +2023-02-07 09:46:02,928 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9794, 2.1948, 1.7324, 2.8696, 1.3813, 1.6339, 2.0823, 2.2504], + device='cuda:0'), covar=tensor([0.0693, 0.0780, 0.0900, 0.0291, 0.1039, 0.1234, 0.0775, 0.0649], + device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0194, 0.0246, 0.0212, 0.0205, 0.0247, 0.0249, 0.0207], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 09:46:11,115 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.69 vs. limit=5.0 +2023-02-07 09:46:12,728 INFO [train.py:901] (0/4) Epoch 26, batch 1300, loss[loss=0.1675, simple_loss=0.2622, pruned_loss=0.03643, over 8516.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2841, pruned_loss=0.05818, over 1619026.22 frames. ], batch size: 28, lr: 2.92e-03, grad_scale: 4.0 +2023-02-07 09:46:29,975 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6820, 1.6608, 2.1077, 1.3140, 1.2935, 2.1176, 0.2559, 1.3491], + device='cuda:0'), covar=tensor([0.1413, 0.1135, 0.0328, 0.1045, 0.2262, 0.0357, 0.1735, 0.1164], + device='cuda:0'), in_proj_covar=tensor([0.0198, 0.0204, 0.0132, 0.0224, 0.0278, 0.0143, 0.0173, 0.0200], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 09:46:31,896 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=203400.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:46:37,059 INFO [optim.py:369] (0/4) 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 09:46:46,424 INFO [train.py:901] (0/4) Epoch 26, batch 1350, loss[loss=0.2235, simple_loss=0.3154, pruned_loss=0.06577, over 8323.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2848, pruned_loss=0.05856, over 1617604.42 frames. ], batch size: 26, lr: 2.92e-03, grad_scale: 4.0 +2023-02-07 09:47:04,178 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=203447.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:47:22,497 INFO [train.py:901] (0/4) Epoch 26, batch 1400, loss[loss=0.1807, simple_loss=0.2752, pruned_loss=0.04307, over 8251.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2855, pruned_loss=0.05891, over 1619106.34 frames. ], batch size: 24, lr: 2.92e-03, grad_scale: 4.0 +2023-02-07 09:47:24,886 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-02-07 09:47:31,497 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 09:47:52,605 INFO [zipformer.py:1185] (0/4) 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] (0/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-07 09:47:57,076 INFO [train.py:901] (0/4) Epoch 26, batch 1450, loss[loss=0.2253, simple_loss=0.2964, pruned_loss=0.07708, over 8086.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2852, pruned_loss=0.05876, over 1622271.21 frames. ], batch size: 21, lr: 2.92e-03, grad_scale: 4.0 +2023-02-07 09:48:24,031 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=203562.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:48:24,749 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203563.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 09:48:31,364 INFO [train.py:901] (0/4) Epoch 26, batch 1500, loss[loss=0.2415, simple_loss=0.3249, pruned_loss=0.07908, over 8502.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.285, pruned_loss=0.05866, over 1624122.77 frames. ], batch size: 26, lr: 2.92e-03, grad_scale: 4.0 +2023-02-07 09:48:39,241 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 +2023-02-07 09:48:42,960 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=203588.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 09:48:56,866 INFO [optim.py:369] (0/4) 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 09:49:06,819 INFO [train.py:901] (0/4) Epoch 26, batch 1550, loss[loss=0.1878, simple_loss=0.285, pruned_loss=0.04533, over 8253.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2854, pruned_loss=0.05911, over 1623495.84 frames. ], batch size: 24, lr: 2.92e-03, grad_scale: 4.0 +2023-02-07 09:49:40,408 INFO [train.py:901] (0/4) Epoch 26, batch 1600, loss[loss=0.2115, simple_loss=0.2945, pruned_loss=0.06424, over 8042.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2859, pruned_loss=0.05947, over 1621287.55 frames. ], batch size: 22, lr: 2.92e-03, grad_scale: 8.0 +2023-02-07 09:50:05,085 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 26, batch 1650, loss[loss=0.1622, simple_loss=0.244, pruned_loss=0.04026, over 7438.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2852, pruned_loss=0.05955, over 1617290.69 frames. ], batch size: 17, lr: 2.92e-03, grad_scale: 8.0 +2023-02-07 09:50:32,730 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.5597, 2.9350, 2.3359, 4.1463, 2.0041, 2.1923, 2.9707, 3.1086], + device='cuda:0'), covar=tensor([0.0750, 0.0741, 0.0880, 0.0231, 0.0961, 0.1204, 0.0769, 0.0751], + device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0195, 0.0247, 0.0213, 0.0205, 0.0248, 0.0250, 0.0207], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 09:50:36,582 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 +2023-02-07 09:50:44,771 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=203765.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:50:45,726 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 +2023-02-07 09:50:48,702 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 26, batch 1700, loss[loss=0.2061, simple_loss=0.2812, pruned_loss=0.0655, over 7799.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2847, pruned_loss=0.05903, over 1617478.59 frames. ], batch size: 20, lr: 2.92e-03, grad_scale: 8.0 +2023-02-07 09:50:57,400 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=203784.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:51:05,466 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=203796.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:51:14,290 INFO [optim.py:369] (0/4) 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 09:51:20,893 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203818.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:51:20,912 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5215, 2.0083, 3.1698, 1.3260, 2.4191, 1.9590, 1.5918, 2.4572], + device='cuda:0'), covar=tensor([0.2355, 0.2966, 0.1020, 0.5518, 0.2185, 0.3773, 0.2894, 0.2483], + device='cuda:0'), in_proj_covar=tensor([0.0534, 0.0624, 0.0554, 0.0659, 0.0654, 0.0602, 0.0552, 0.0639], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 09:51:23,311 INFO [train.py:901] (0/4) Epoch 26, batch 1750, loss[loss=0.2067, simple_loss=0.2923, pruned_loss=0.06059, over 8518.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2859, pruned_loss=0.05948, over 1622620.40 frames. ], batch size: 26, lr: 2.92e-03, grad_scale: 8.0 +2023-02-07 09:51:28,096 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=203829.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:51:38,338 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=203843.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:51:39,327 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 +2023-02-07 09:51:43,135 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4536, 1.4217, 1.8269, 1.1647, 1.0521, 1.7981, 0.1489, 1.1483], + device='cuda:0'), covar=tensor([0.1385, 0.1346, 0.0413, 0.0923, 0.2599, 0.0495, 0.1902, 0.1217], + device='cuda:0'), in_proj_covar=tensor([0.0198, 0.0205, 0.0133, 0.0226, 0.0280, 0.0144, 0.0173, 0.0199], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 09:51:45,962 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0708, 2.2854, 3.1147, 1.9423, 2.6539, 2.3160, 2.1302, 2.6099], + device='cuda:0'), covar=tensor([0.1518, 0.2111, 0.0720, 0.3544, 0.1420, 0.2450, 0.1807, 0.1904], + device='cuda:0'), in_proj_covar=tensor([0.0535, 0.0625, 0.0555, 0.0660, 0.0655, 0.0603, 0.0552, 0.0640], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 09:51:54,043 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=203865.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:51:58,807 INFO [train.py:901] (0/4) Epoch 26, batch 1800, loss[loss=0.1925, simple_loss=0.2813, pruned_loss=0.05188, over 8487.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2849, pruned_loss=0.0593, over 1620513.93 frames. ], batch size: 29, lr: 2.92e-03, grad_scale: 8.0 +2023-02-07 09:52:00,429 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8052, 2.4859, 3.9384, 1.6161, 2.8983, 2.2128, 1.9849, 2.8915], + device='cuda:0'), covar=tensor([0.2166, 0.2772, 0.1066, 0.5434, 0.2161, 0.3756, 0.2784, 0.2855], + device='cuda:0'), in_proj_covar=tensor([0.0534, 0.0623, 0.0554, 0.0659, 0.0655, 0.0603, 0.0551, 0.0639], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 09:52:21,984 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4120, 1.6045, 2.1491, 1.3230, 1.3689, 1.6707, 1.4148, 1.5807], + device='cuda:0'), covar=tensor([0.1927, 0.2593, 0.0978, 0.4642, 0.2210, 0.3394, 0.2461, 0.2083], + device='cuda:0'), in_proj_covar=tensor([0.0534, 0.0623, 0.0554, 0.0659, 0.0655, 0.0603, 0.0551, 0.0639], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 09:52:23,002 INFO [optim.py:369] (0/4) 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 09:52:29,699 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=203918.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 09:52:32,264 INFO [train.py:901] (0/4) Epoch 26, batch 1850, loss[loss=0.2551, simple_loss=0.3177, pruned_loss=0.09627, over 6846.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2848, pruned_loss=0.05959, over 1620270.13 frames. ], batch size: 71, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 09:52:47,698 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=203944.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:53:07,698 INFO [train.py:901] (0/4) Epoch 26, batch 1900, loss[loss=0.2174, simple_loss=0.3059, pruned_loss=0.06448, over 8657.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2839, pruned_loss=0.05899, over 1617597.04 frames. ], batch size: 34, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 09:53:27,230 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-204000.pt +2023-02-07 09:53:33,464 INFO [optim.py:369] (0/4) 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 09:53:36,913 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-07 09:53:42,920 INFO [train.py:901] (0/4) Epoch 26, batch 1950, loss[loss=0.2311, simple_loss=0.3115, pruned_loss=0.07534, over 8561.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.283, pruned_loss=0.05841, over 1617015.50 frames. ], batch size: 49, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 09:53:49,445 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-07 09:53:57,074 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.6424, 3.5789, 3.3567, 2.2207, 3.2508, 3.3127, 3.3170, 3.1963], + device='cuda:0'), covar=tensor([0.0849, 0.0671, 0.0933, 0.3818, 0.0860, 0.1200, 0.1301, 0.0902], + device='cuda:0'), in_proj_covar=tensor([0.0538, 0.0454, 0.0438, 0.0554, 0.0438, 0.0458, 0.0434, 0.0397], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 09:54:07,155 WARNING [train.py:1067] (0/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-07 09:54:17,200 INFO [train.py:901] (0/4) Epoch 26, batch 2000, loss[loss=0.1803, simple_loss=0.2488, pruned_loss=0.05589, over 7449.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2837, pruned_loss=0.05874, over 1614762.87 frames. ], batch size: 17, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 09:54:34,351 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 09:54:44,458 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=204109.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:54:53,075 INFO [train.py:901] (0/4) Epoch 26, batch 2050, loss[loss=0.2032, simple_loss=0.2739, pruned_loss=0.06626, over 8131.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.284, pruned_loss=0.05876, over 1618351.86 frames. ], batch size: 22, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 09:54:57,136 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=204128.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:55:15,938 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3707, 1.4424, 1.3885, 1.8277, 0.7708, 1.2713, 1.3118, 1.4498], + device='cuda:0'), covar=tensor([0.0849, 0.0739, 0.0896, 0.0467, 0.1080, 0.1330, 0.0721, 0.0687], + device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0194, 0.0245, 0.0212, 0.0205, 0.0247, 0.0249, 0.0205], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 09:55:26,603 INFO [train.py:901] (0/4) Epoch 26, batch 2100, loss[loss=0.1691, simple_loss=0.258, pruned_loss=0.04016, over 8087.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2848, pruned_loss=0.05891, over 1619518.27 frames. ], batch size: 21, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 09:55:33,651 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=204181.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:55:47,793 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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] (0/4) 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] (0/4) Epoch 26, batch 2150, loss[loss=0.2471, simple_loss=0.3135, pruned_loss=0.09029, over 7041.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2846, pruned_loss=0.05938, over 1616830.07 frames. ], batch size: 71, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 09:56:03,978 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204224.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:56:04,613 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=204225.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:56:04,835 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.60 vs. limit=2.0 +2023-02-07 09:56:17,181 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204243.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:56:29,888 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=204262.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 09:56:36,278 INFO [train.py:901] (0/4) Epoch 26, batch 2200, loss[loss=0.1811, simple_loss=0.254, pruned_loss=0.05416, over 7430.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2847, pruned_loss=0.0593, over 1616871.50 frames. ], batch size: 17, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 09:57:01,395 INFO [optim.py:369] (0/4) 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 09:57:11,826 INFO [train.py:901] (0/4) Epoch 26, batch 2250, loss[loss=0.2032, simple_loss=0.2801, pruned_loss=0.06315, over 8471.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2854, pruned_loss=0.05968, over 1614221.19 frames. ], batch size: 27, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 09:57:13,359 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204324.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:57:46,485 INFO [train.py:901] (0/4) Epoch 26, batch 2300, loss[loss=0.2028, simple_loss=0.292, pruned_loss=0.05674, over 8108.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2846, pruned_loss=0.05885, over 1615570.01 frames. ], batch size: 23, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 09:57:50,096 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204377.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 09:57:57,430 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1169, 2.4784, 2.7889, 1.4820, 3.0074, 1.7904, 1.4529, 2.2379], + device='cuda:0'), covar=tensor([0.0949, 0.0513, 0.0389, 0.0997, 0.0518, 0.1053, 0.1037, 0.0653], + device='cuda:0'), in_proj_covar=tensor([0.0467, 0.0405, 0.0359, 0.0456, 0.0392, 0.0547, 0.0400, 0.0434], + device='cuda:0'), out_proj_covar=tensor([1.2417e-04, 1.0545e-04, 9.3790e-05, 1.1938e-04, 1.0260e-04, 1.5299e-04, + 1.0700e-04, 1.1407e-04], device='cuda:0') +2023-02-07 09:58:10,725 INFO [optim.py:369] (0/4) 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 09:58:17,209 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6714, 1.9873, 2.0817, 1.3743, 2.0746, 1.5461, 0.5405, 1.9418], + device='cuda:0'), covar=tensor([0.0647, 0.0373, 0.0297, 0.0607, 0.0450, 0.1005, 0.1040, 0.0296], + device='cuda:0'), in_proj_covar=tensor([0.0466, 0.0404, 0.0359, 0.0455, 0.0391, 0.0546, 0.0399, 0.0433], + device='cuda:0'), out_proj_covar=tensor([1.2390e-04, 1.0532e-04, 9.3736e-05, 1.1916e-04, 1.0234e-04, 1.5264e-04, + 1.0674e-04, 1.1402e-04], device='cuda:0') +2023-02-07 09:58:21,053 INFO [train.py:901] (0/4) Epoch 26, batch 2350, loss[loss=0.1849, simple_loss=0.275, pruned_loss=0.04741, over 7822.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2846, pruned_loss=0.05882, over 1616617.33 frames. ], batch size: 20, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 09:58:33,848 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=204439.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:58:57,206 INFO [train.py:901] (0/4) Epoch 26, batch 2400, loss[loss=0.2032, simple_loss=0.2988, pruned_loss=0.05383, over 8477.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.285, pruned_loss=0.05942, over 1615233.86 frames. ], batch size: 25, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 09:59:02,850 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=204480.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:59:16,083 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=204499.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:59:20,382 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=204505.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:59:22,321 INFO [optim.py:369] (0/4) 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 09:59:32,143 INFO [train.py:901] (0/4) Epoch 26, batch 2450, loss[loss=0.2382, simple_loss=0.3143, pruned_loss=0.08104, over 8347.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2848, pruned_loss=0.05973, over 1614325.91 frames. ], batch size: 26, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 09:59:33,753 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=204524.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:59:34,321 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=204525.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 09:59:56,587 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 26, batch 2500, loss[loss=0.1955, simple_loss=0.2827, pruned_loss=0.05411, over 8605.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2849, pruned_loss=0.05953, over 1615067.58 frames. ], batch size: 39, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 10:00:15,144 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=204580.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:00:25,197 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6594, 2.6423, 1.8877, 2.4829, 2.2362, 1.6605, 2.2585, 2.2917], + device='cuda:0'), covar=tensor([0.1672, 0.0464, 0.1309, 0.0668, 0.0778, 0.1652, 0.1056, 0.1090], + device='cuda:0'), in_proj_covar=tensor([0.0358, 0.0240, 0.0341, 0.0312, 0.0303, 0.0346, 0.0348, 0.0321], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 10:00:31,996 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=204605.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:00:33,809 INFO [optim.py:369] (0/4) 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 10:00:43,171 INFO [train.py:901] (0/4) Epoch 26, batch 2550, loss[loss=0.1717, simple_loss=0.2492, pruned_loss=0.04712, over 7434.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2841, pruned_loss=0.05912, over 1611694.83 frames. ], batch size: 17, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 10:00:50,563 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=204633.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 10:00:55,139 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204640.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:01:07,859 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=204658.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 10:01:18,432 INFO [train.py:901] (0/4) Epoch 26, batch 2600, loss[loss=0.186, simple_loss=0.2655, pruned_loss=0.05324, over 7520.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2841, pruned_loss=0.0589, over 1615656.99 frames. ], batch size: 18, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 10:01:43,337 INFO [optim.py:369] (0/4) 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 10:01:52,890 INFO [train.py:901] (0/4) Epoch 26, batch 2650, loss[loss=0.1698, simple_loss=0.2454, pruned_loss=0.04707, over 7712.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2855, pruned_loss=0.05973, over 1618225.86 frames. ], batch size: 18, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 10:02:14,671 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 +2023-02-07 10:02:28,033 INFO [train.py:901] (0/4) Epoch 26, batch 2700, loss[loss=0.2164, simple_loss=0.3102, pruned_loss=0.06128, over 8354.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2851, pruned_loss=0.05952, over 1614379.19 frames. ], batch size: 26, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 10:02:53,800 INFO [optim.py:369] (0/4) 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 10:02:55,422 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=204810.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:03:04,008 INFO [train.py:901] (0/4) Epoch 26, batch 2750, loss[loss=0.2275, simple_loss=0.3246, pruned_loss=0.06526, over 8465.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2851, pruned_loss=0.05918, over 1617702.17 frames. ], batch size: 25, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 10:03:13,288 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=204835.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:03:14,638 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0707, 1.3725, 1.6706, 1.2560, 0.7467, 1.4542, 1.1880, 1.0336], + device='cuda:0'), covar=tensor([0.0627, 0.1200, 0.1591, 0.1402, 0.0556, 0.1429, 0.0656, 0.0719], + device='cuda:0'), in_proj_covar=tensor([0.0098, 0.0153, 0.0190, 0.0161, 0.0100, 0.0163, 0.0112, 0.0145], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:0') +2023-02-07 10:03:16,002 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=204839.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:03:38,122 INFO [train.py:901] (0/4) Epoch 26, batch 2800, loss[loss=0.2401, simple_loss=0.3163, pruned_loss=0.08197, over 8377.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2842, pruned_loss=0.05886, over 1615988.06 frames. ], batch size: 49, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 10:03:53,216 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5914, 1.2694, 2.8772, 1.2731, 2.1319, 3.0922, 3.2229, 2.6218], + device='cuda:0'), covar=tensor([0.1336, 0.1880, 0.0397, 0.2261, 0.0917, 0.0302, 0.0692, 0.0622], + device='cuda:0'), in_proj_covar=tensor([0.0304, 0.0328, 0.0290, 0.0320, 0.0322, 0.0276, 0.0436, 0.0308], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 10:03:55,282 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=204896.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:04:03,030 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.78 vs. limit=5.0 +2023-02-07 10:04:04,626 INFO [optim.py:369] (0/4) 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 10:04:13,906 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=204921.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:04:14,413 INFO [train.py:901] (0/4) Epoch 26, batch 2850, loss[loss=0.1872, simple_loss=0.2573, pruned_loss=0.05857, over 7426.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2841, pruned_loss=0.05861, over 1616069.43 frames. ], batch size: 17, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 10:04:48,524 INFO [train.py:901] (0/4) Epoch 26, batch 2900, loss[loss=0.2229, simple_loss=0.3036, pruned_loss=0.07109, over 8493.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2855, pruned_loss=0.05944, over 1614366.82 frames. ], batch size: 26, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 10:05:13,383 INFO [optim.py:369] (0/4) 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 10:05:20,488 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 26, batch 2950, loss[loss=0.258, simple_loss=0.3299, pruned_loss=0.09302, over 8434.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2861, pruned_loss=0.06002, over 1616069.85 frames. ], batch size: 27, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 10:05:52,963 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 +2023-02-07 10:05:58,486 INFO [train.py:901] (0/4) Epoch 26, batch 3000, loss[loss=0.2005, simple_loss=0.2787, pruned_loss=0.06112, over 8335.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2863, pruned_loss=0.05966, over 1612946.14 frames. ], batch size: 26, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 10:05:58,487 INFO [train.py:926] (0/4) Computing validation loss +2023-02-07 10:06:11,416 INFO [train.py:935] (0/4) Epoch 26, validation: loss=0.1716, simple_loss=0.2713, pruned_loss=0.03593, over 944034.00 frames. +2023-02-07 10:06:11,417 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6760MB +2023-02-07 10:06:36,696 INFO [optim.py:369] (0/4) 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 10:06:45,997 INFO [train.py:901] (0/4) Epoch 26, batch 3050, loss[loss=0.1909, simple_loss=0.2718, pruned_loss=0.05499, over 7965.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2866, pruned_loss=0.06014, over 1616177.42 frames. ], batch size: 21, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 10:06:47,781 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 +2023-02-07 10:06:49,433 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=205127.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:07:22,789 INFO [train.py:901] (0/4) Epoch 26, batch 3100, loss[loss=0.1763, simple_loss=0.2668, pruned_loss=0.04295, over 8092.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2865, pruned_loss=0.06009, over 1621382.13 frames. ], batch size: 21, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 10:07:24,530 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.78 vs. limit=5.0 +2023-02-07 10:07:30,228 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=205183.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:07:48,156 INFO [optim.py:369] (0/4) 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 10:07:53,683 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=205216.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:07:57,550 INFO [train.py:901] (0/4) Epoch 26, batch 3150, loss[loss=0.1824, simple_loss=0.2674, pruned_loss=0.0487, over 8063.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2867, pruned_loss=0.06016, over 1624167.30 frames. ], batch size: 21, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 10:08:33,477 INFO [train.py:901] (0/4) Epoch 26, batch 3200, loss[loss=0.1802, simple_loss=0.2632, pruned_loss=0.04863, over 7812.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2845, pruned_loss=0.0591, over 1621126.77 frames. ], batch size: 20, lr: 2.91e-03, grad_scale: 8.0 +2023-02-07 10:08:52,246 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=205298.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:08:52,270 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7589, 2.6244, 1.8883, 2.3245, 2.2993, 1.6618, 2.1724, 2.2166], + device='cuda:0'), covar=tensor([0.1362, 0.0422, 0.1216, 0.0620, 0.0636, 0.1474, 0.1000, 0.0916], + device='cuda:0'), in_proj_covar=tensor([0.0356, 0.0238, 0.0337, 0.0309, 0.0299, 0.0341, 0.0344, 0.0318], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 10:08:58,807 INFO [optim.py:369] (0/4) 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 10:08:58,936 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.8391, 3.7769, 3.4660, 1.9162, 3.3886, 3.5080, 3.4101, 3.3390], + device='cuda:0'), covar=tensor([0.0897, 0.0628, 0.1209, 0.4570, 0.0968, 0.1074, 0.1404, 0.0843], + device='cuda:0'), in_proj_covar=tensor([0.0534, 0.0449, 0.0437, 0.0549, 0.0432, 0.0455, 0.0431, 0.0394], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 10:09:09,092 INFO [train.py:901] (0/4) Epoch 26, batch 3250, loss[loss=0.2069, simple_loss=0.2918, pruned_loss=0.06101, over 8293.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.285, pruned_loss=0.05928, over 1623959.20 frames. ], batch size: 23, lr: 2.90e-03, grad_scale: 16.0 +2023-02-07 10:09:39,876 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.2192, 5.2444, 4.7364, 2.8939, 4.6263, 4.9160, 4.8585, 4.6707], + device='cuda:0'), covar=tensor([0.0576, 0.0433, 0.0856, 0.3746, 0.0874, 0.0919, 0.1051, 0.0679], + device='cuda:0'), in_proj_covar=tensor([0.0531, 0.0448, 0.0435, 0.0547, 0.0431, 0.0453, 0.0429, 0.0393], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 10:09:43,186 INFO [train.py:901] (0/4) Epoch 26, batch 3300, loss[loss=0.1579, simple_loss=0.2388, pruned_loss=0.03849, over 7904.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2857, pruned_loss=0.05971, over 1622942.70 frames. ], batch size: 20, lr: 2.90e-03, grad_scale: 16.0 +2023-02-07 10:10:10,314 INFO [optim.py:369] (0/4) 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 10:10:20,066 INFO [train.py:901] (0/4) Epoch 26, batch 3350, loss[loss=0.2318, simple_loss=0.315, pruned_loss=0.07433, over 8594.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2852, pruned_loss=0.05909, over 1623908.13 frames. ], batch size: 34, lr: 2.90e-03, grad_scale: 16.0 +2023-02-07 10:10:32,196 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 +2023-02-07 10:10:54,100 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=205471.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:10:54,722 INFO [train.py:901] (0/4) Epoch 26, batch 3400, loss[loss=0.2276, simple_loss=0.3161, pruned_loss=0.06958, over 8398.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2839, pruned_loss=0.05838, over 1624103.50 frames. ], batch size: 49, lr: 2.90e-03, grad_scale: 16.0 +2023-02-07 10:11:07,754 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 +2023-02-07 10:11:20,313 INFO [optim.py:369] (0/4) 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 10:11:30,475 INFO [train.py:901] (0/4) Epoch 26, batch 3450, loss[loss=0.203, simple_loss=0.2857, pruned_loss=0.06012, over 8526.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2834, pruned_loss=0.05804, over 1620321.80 frames. ], batch size: 49, lr: 2.90e-03, grad_scale: 16.0 +2023-02-07 10:11:45,457 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7365, 1.3239, 2.8904, 1.4650, 2.2529, 3.0941, 3.2321, 2.6478], + device='cuda:0'), covar=tensor([0.1168, 0.1813, 0.0353, 0.2040, 0.0866, 0.0293, 0.0595, 0.0599], + device='cuda:0'), in_proj_covar=tensor([0.0302, 0.0326, 0.0289, 0.0318, 0.0320, 0.0275, 0.0434, 0.0307], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 10:11:53,149 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=205554.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:11:54,048 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.96 vs. limit=5.0 +2023-02-07 10:11:56,802 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 +2023-02-07 10:11:57,155 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=205560.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:12:05,241 INFO [train.py:901] (0/4) Epoch 26, batch 3500, loss[loss=0.1962, simple_loss=0.2875, pruned_loss=0.05242, over 8481.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2827, pruned_loss=0.05775, over 1618425.88 frames. ], batch size: 25, lr: 2.90e-03, grad_scale: 16.0 +2023-02-07 10:12:10,353 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=205579.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:12:15,144 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=205586.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:12:24,883 WARNING [train.py:1067] (0/4) 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] (0/4) 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 10:12:39,510 INFO [train.py:901] (0/4) Epoch 26, batch 3550, loss[loss=0.2119, simple_loss=0.2993, pruned_loss=0.06224, over 8108.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2834, pruned_loss=0.05795, over 1617514.28 frames. ], batch size: 23, lr: 2.90e-03, grad_scale: 16.0 +2023-02-07 10:12:59,247 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 +2023-02-07 10:13:08,106 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.3819, 2.7113, 3.0498, 1.8065, 3.4335, 2.0904, 1.6271, 2.3927], + device='cuda:0'), covar=tensor([0.0765, 0.0400, 0.0282, 0.0841, 0.0381, 0.0847, 0.1063, 0.0542], + device='cuda:0'), in_proj_covar=tensor([0.0470, 0.0407, 0.0363, 0.0458, 0.0391, 0.0553, 0.0402, 0.0436], + device='cuda:0'), out_proj_covar=tensor([1.2482e-04, 1.0608e-04, 9.4851e-05, 1.1996e-04, 1.0242e-04, 1.5480e-04, + 1.0754e-04, 1.1466e-04], device='cuda:0') +2023-02-07 10:13:12,849 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7727, 1.9970, 1.6009, 2.7833, 1.2870, 1.4448, 1.9742, 2.1859], + device='cuda:0'), covar=tensor([0.0906, 0.0937, 0.1141, 0.0402, 0.1181, 0.1488, 0.0905, 0.0772], + device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0193, 0.0244, 0.0210, 0.0202, 0.0245, 0.0247, 0.0204], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 10:13:14,225 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7731, 2.0325, 2.9862, 1.5923, 2.3592, 2.2085, 1.7359, 2.3780], + device='cuda:0'), covar=tensor([0.1796, 0.2538, 0.0837, 0.4360, 0.1659, 0.3008, 0.2336, 0.1995], + device='cuda:0'), in_proj_covar=tensor([0.0536, 0.0625, 0.0558, 0.0662, 0.0657, 0.0604, 0.0557, 0.0640], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 10:13:15,380 INFO [train.py:901] (0/4) Epoch 26, batch 3600, loss[loss=0.1603, simple_loss=0.2432, pruned_loss=0.03869, over 7316.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2835, pruned_loss=0.05835, over 1617163.30 frames. ], batch size: 16, lr: 2.90e-03, grad_scale: 16.0 +2023-02-07 10:13:16,205 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=205673.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:13:17,471 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=205675.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:13:21,537 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0356, 1.6113, 3.3671, 1.5154, 2.4495, 3.7205, 3.8522, 3.1196], + device='cuda:0'), covar=tensor([0.1222, 0.1813, 0.0351, 0.2178, 0.1073, 0.0246, 0.0531, 0.0592], + device='cuda:0'), in_proj_covar=tensor([0.0304, 0.0328, 0.0291, 0.0320, 0.0322, 0.0277, 0.0437, 0.0308], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 10:13:37,463 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 +2023-02-07 10:13:39,662 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 26, batch 3650, loss[loss=0.1651, simple_loss=0.2453, pruned_loss=0.04243, over 7641.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.283, pruned_loss=0.05804, over 1612244.72 frames. ], batch size: 19, lr: 2.90e-03, grad_scale: 16.0 +2023-02-07 10:13:56,158 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8001, 1.7578, 2.3252, 1.4674, 1.3206, 2.3052, 0.3406, 1.4244], + device='cuda:0'), covar=tensor([0.1578, 0.1109, 0.0291, 0.1045, 0.2291, 0.0306, 0.1930, 0.1191], + device='cuda:0'), in_proj_covar=tensor([0.0195, 0.0201, 0.0131, 0.0220, 0.0274, 0.0143, 0.0170, 0.0197], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 10:14:18,442 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 26, batch 3700, loss[loss=0.1543, simple_loss=0.2419, pruned_loss=0.03338, over 7791.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2825, pruned_loss=0.05808, over 1610864.72 frames. ], batch size: 19, lr: 2.90e-03, grad_scale: 16.0 +2023-02-07 10:14:27,591 WARNING [train.py:1067] (0/4) 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] (0/4) 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] (0/4) Epoch 26, batch 3750, loss[loss=0.2099, simple_loss=0.291, pruned_loss=0.06441, over 8612.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.282, pruned_loss=0.05783, over 1607224.28 frames. ], batch size: 31, lr: 2.90e-03, grad_scale: 16.0 +2023-02-07 10:15:12,916 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=205842.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:15:14,925 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6971, 1.8961, 1.9988, 1.4833, 2.1761, 1.5426, 0.8163, 1.8955], + device='cuda:0'), covar=tensor([0.0677, 0.0394, 0.0338, 0.0612, 0.0436, 0.0930, 0.0936, 0.0352], + device='cuda:0'), in_proj_covar=tensor([0.0470, 0.0408, 0.0362, 0.0458, 0.0391, 0.0552, 0.0402, 0.0436], + device='cuda:0'), out_proj_covar=tensor([1.2485e-04, 1.0618e-04, 9.4536e-05, 1.1998e-04, 1.0244e-04, 1.5464e-04, + 1.0771e-04, 1.1476e-04], device='cuda:0') +2023-02-07 10:15:24,381 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0541, 3.6302, 2.3024, 2.9716, 2.7234, 2.1079, 2.7570, 3.1115], + device='cuda:0'), covar=tensor([0.1837, 0.0353, 0.1226, 0.0734, 0.0820, 0.1469, 0.1194, 0.1046], + device='cuda:0'), in_proj_covar=tensor([0.0358, 0.0239, 0.0341, 0.0311, 0.0302, 0.0344, 0.0347, 0.0322], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 10:15:31,317 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=205867.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:15:34,438 INFO [train.py:901] (0/4) Epoch 26, batch 3800, loss[loss=0.2197, simple_loss=0.3085, pruned_loss=0.06545, over 8239.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2814, pruned_loss=0.05699, over 1610900.62 frames. ], batch size: 24, lr: 2.90e-03, grad_scale: 16.0 +2023-02-07 10:15:52,886 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4817, 2.4330, 3.1906, 2.5188, 3.1513, 2.5224, 2.4651, 2.0179], + device='cuda:0'), covar=tensor([0.5831, 0.5123, 0.1997, 0.4070, 0.2536, 0.3155, 0.1980, 0.5369], + device='cuda:0'), in_proj_covar=tensor([0.0956, 0.1005, 0.0825, 0.0981, 0.1017, 0.0919, 0.0764, 0.0841], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 10:15:59,216 INFO [optim.py:369] (0/4) 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 10:15:59,374 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=205908.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:16:08,813 INFO [train.py:901] (0/4) Epoch 26, batch 3850, loss[loss=0.2021, simple_loss=0.2964, pruned_loss=0.05383, over 8505.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2818, pruned_loss=0.05727, over 1609632.99 frames. ], batch size: 26, lr: 2.90e-03, grad_scale: 16.0 +2023-02-07 10:16:15,133 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=205931.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:16:29,197 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-07 10:16:29,384 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5569, 2.6130, 1.8273, 2.4029, 2.2010, 1.6377, 2.1472, 2.2245], + device='cuda:0'), covar=tensor([0.1683, 0.0403, 0.1245, 0.0695, 0.0839, 0.1594, 0.1079, 0.1116], + device='cuda:0'), in_proj_covar=tensor([0.0355, 0.0238, 0.0337, 0.0309, 0.0300, 0.0342, 0.0345, 0.0320], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 10:16:32,105 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=205956.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:16:36,156 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4724, 1.2537, 2.3723, 1.3533, 2.2194, 2.5298, 2.7024, 2.1796], + device='cuda:0'), covar=tensor([0.1111, 0.1512, 0.0397, 0.2012, 0.0767, 0.0377, 0.0648, 0.0603], + device='cuda:0'), in_proj_covar=tensor([0.0304, 0.0328, 0.0291, 0.0321, 0.0322, 0.0278, 0.0438, 0.0309], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 10:16:42,784 INFO [train.py:901] (0/4) Epoch 26, batch 3900, loss[loss=0.2037, simple_loss=0.2998, pruned_loss=0.05385, over 8464.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2823, pruned_loss=0.05732, over 1612454.74 frames. ], batch size: 27, lr: 2.90e-03, grad_scale: 16.0 +2023-02-07 10:17:03,801 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-206000.pt +2023-02-07 10:17:05,680 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.02 vs. limit=5.0 +2023-02-07 10:17:09,988 INFO [optim.py:369] (0/4) 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 10:17:12,781 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4200, 2.0278, 3.0031, 1.7796, 1.5568, 2.9880, 0.7737, 2.1951], + device='cuda:0'), covar=tensor([0.1156, 0.1231, 0.0300, 0.1243, 0.2406, 0.0297, 0.1935, 0.1142], + device='cuda:0'), in_proj_covar=tensor([0.0195, 0.0200, 0.0130, 0.0219, 0.0272, 0.0143, 0.0170, 0.0196], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 10:17:16,697 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=206017.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:17:20,037 INFO [train.py:901] (0/4) Epoch 26, batch 3950, loss[loss=0.1887, simple_loss=0.2855, pruned_loss=0.04594, over 8464.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2825, pruned_loss=0.05722, over 1615385.69 frames. ], batch size: 25, lr: 2.90e-03, grad_scale: 16.0 +2023-02-07 10:17:49,442 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0611, 1.8263, 2.3169, 2.0138, 2.3170, 2.1352, 1.9754, 1.0845], + device='cuda:0'), covar=tensor([0.5768, 0.4772, 0.2093, 0.3770, 0.2559, 0.3168, 0.1972, 0.5269], + device='cuda:0'), in_proj_covar=tensor([0.0955, 0.1004, 0.0826, 0.0978, 0.1017, 0.0918, 0.0762, 0.0840], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 10:17:53,924 INFO [train.py:901] (0/4) Epoch 26, batch 4000, loss[loss=0.1777, simple_loss=0.2711, pruned_loss=0.04218, over 7228.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2821, pruned_loss=0.05721, over 1615548.28 frames. ], batch size: 16, lr: 2.90e-03, grad_scale: 16.0 +2023-02-07 10:17:55,452 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=206074.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:18:18,670 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 10:18:29,518 INFO [train.py:901] (0/4) Epoch 26, batch 4050, loss[loss=0.2111, simple_loss=0.3017, pruned_loss=0.0602, over 8323.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2832, pruned_loss=0.05819, over 1615821.41 frames. ], batch size: 25, lr: 2.90e-03, grad_scale: 16.0 +2023-02-07 10:18:37,175 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=206132.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:18:52,053 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3465, 1.6869, 1.2933, 2.6510, 1.3023, 1.2316, 1.9802, 1.8495], + device='cuda:0'), covar=tensor([0.1530, 0.1155, 0.1919, 0.0375, 0.1173, 0.2003, 0.0778, 0.0950], + device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0192, 0.0242, 0.0210, 0.0201, 0.0244, 0.0246, 0.0203], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 10:19:03,799 INFO [train.py:901] (0/4) Epoch 26, batch 4100, loss[loss=0.1928, simple_loss=0.2771, pruned_loss=0.05427, over 8594.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2837, pruned_loss=0.05845, over 1620160.71 frames. ], batch size: 34, lr: 2.90e-03, grad_scale: 16.0 +2023-02-07 10:19:28,869 INFO [optim.py:369] (0/4) 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 10:19:34,505 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=206215.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:19:39,248 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 26, batch 4150, loss[loss=0.213, simple_loss=0.28, pruned_loss=0.07305, over 7933.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2834, pruned_loss=0.05817, over 1621701.39 frames. ], batch size: 20, lr: 2.90e-03, grad_scale: 16.0 +2023-02-07 10:20:00,781 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=206252.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:20:14,207 INFO [train.py:901] (0/4) Epoch 26, batch 4200, loss[loss=0.2055, simple_loss=0.2963, pruned_loss=0.05737, over 8368.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2845, pruned_loss=0.05857, over 1622058.88 frames. ], batch size: 24, lr: 2.90e-03, grad_scale: 16.0 +2023-02-07 10:20:22,976 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-07 10:20:38,388 INFO [optim.py:369] (0/4) 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 10:20:44,915 WARNING [train.py:1067] (0/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-07 10:20:49,126 INFO [train.py:901] (0/4) Epoch 26, batch 4250, loss[loss=0.1719, simple_loss=0.2504, pruned_loss=0.04674, over 7697.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2835, pruned_loss=0.05815, over 1618472.04 frames. ], batch size: 18, lr: 2.90e-03, grad_scale: 8.0 +2023-02-07 10:20:54,128 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7151, 1.5021, 2.9125, 1.4202, 2.3123, 3.1437, 3.2421, 2.6885], + device='cuda:0'), covar=tensor([0.1151, 0.1605, 0.0354, 0.2069, 0.0799, 0.0290, 0.0594, 0.0559], + device='cuda:0'), in_proj_covar=tensor([0.0302, 0.0325, 0.0290, 0.0319, 0.0320, 0.0277, 0.0435, 0.0308], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 10:21:21,335 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=206367.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:21:24,650 INFO [train.py:901] (0/4) Epoch 26, batch 4300, loss[loss=0.15, simple_loss=0.2429, pruned_loss=0.02857, over 6764.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2825, pruned_loss=0.0577, over 1619522.55 frames. ], batch size: 15, lr: 2.90e-03, grad_scale: 8.0 +2023-02-07 10:21:35,899 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=206388.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:21:50,296 INFO [optim.py:369] (0/4) 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 10:21:53,271 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=206413.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:21:56,610 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=206418.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:21:59,355 INFO [train.py:901] (0/4) Epoch 26, batch 4350, loss[loss=0.1966, simple_loss=0.2683, pruned_loss=0.06248, over 7813.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.283, pruned_loss=0.05812, over 1618080.31 frames. ], batch size: 20, lr: 2.90e-03, grad_scale: 8.0 +2023-02-07 10:22:02,343 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([0.9663, 1.1879, 1.0848, 1.9271, 0.8250, 0.9684, 1.4217, 1.3432], + device='cuda:0'), covar=tensor([0.1675, 0.1171, 0.1906, 0.0473, 0.1219, 0.2017, 0.0729, 0.0953], + device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0194, 0.0244, 0.0212, 0.0203, 0.0247, 0.0250, 0.0205], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 10:22:18,981 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-07 10:22:34,786 INFO [train.py:901] (0/4) Epoch 26, batch 4400, loss[loss=0.1889, simple_loss=0.2564, pruned_loss=0.06065, over 7441.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2839, pruned_loss=0.05873, over 1618678.47 frames. ], batch size: 17, lr: 2.90e-03, grad_scale: 8.0 +2023-02-07 10:22:38,336 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=206477.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:22:55,676 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=206502.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:23:00,161 INFO [optim.py:369] (0/4) 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 10:23:00,194 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-07 10:23:08,805 INFO [train.py:901] (0/4) Epoch 26, batch 4450, loss[loss=0.1999, simple_loss=0.2892, pruned_loss=0.05527, over 8462.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2846, pruned_loss=0.05923, over 1620383.21 frames. ], batch size: 27, lr: 2.90e-03, grad_scale: 8.0 +2023-02-07 10:23:16,246 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=206533.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:23:34,134 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=206559.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:23:44,257 INFO [train.py:901] (0/4) Epoch 26, batch 4500, loss[loss=0.2005, simple_loss=0.2909, pruned_loss=0.05512, over 8508.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2842, pruned_loss=0.05886, over 1623698.87 frames. ], batch size: 26, lr: 2.90e-03, grad_scale: 8.0 +2023-02-07 10:23:55,214 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-07 10:24:05,596 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9847, 1.5681, 1.8212, 1.3881, 1.0523, 1.5553, 1.8097, 1.5222], + device='cuda:0'), covar=tensor([0.0533, 0.1278, 0.1602, 0.1480, 0.0619, 0.1487, 0.0689, 0.0688], + device='cuda:0'), in_proj_covar=tensor([0.0098, 0.0153, 0.0190, 0.0161, 0.0101, 0.0163, 0.0113, 0.0146], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:0') +2023-02-07 10:24:10,065 INFO [optim.py:369] (0/4) 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 10:24:18,680 INFO [train.py:901] (0/4) Epoch 26, batch 4550, loss[loss=0.2143, simple_loss=0.3029, pruned_loss=0.06289, over 8322.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2826, pruned_loss=0.05793, over 1619054.44 frames. ], batch size: 26, lr: 2.90e-03, grad_scale: 8.0 +2023-02-07 10:24:19,481 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=206623.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:24:35,787 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=206648.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:24:51,533 INFO [train.py:901] (0/4) Epoch 26, batch 4600, loss[loss=0.1764, simple_loss=0.2682, pruned_loss=0.04232, over 7796.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.284, pruned_loss=0.05888, over 1617268.64 frames. ], batch size: 20, lr: 2.90e-03, grad_scale: 8.0 +2023-02-07 10:24:53,052 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=206674.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:25:06,512 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0417, 1.2573, 1.1847, 0.7088, 1.2200, 1.1023, 0.0702, 1.2154], + device='cuda:0'), covar=tensor([0.0455, 0.0422, 0.0394, 0.0648, 0.0468, 0.0931, 0.0933, 0.0381], + device='cuda:0'), in_proj_covar=tensor([0.0468, 0.0405, 0.0362, 0.0458, 0.0391, 0.0550, 0.0401, 0.0437], + device='cuda:0'), out_proj_covar=tensor([1.2427e-04, 1.0548e-04, 9.4578e-05, 1.2001e-04, 1.0246e-04, 1.5383e-04, + 1.0739e-04, 1.1479e-04], device='cuda:0') +2023-02-07 10:25:16,709 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7859, 1.5710, 2.8812, 1.3324, 2.2530, 3.0585, 3.2368, 2.6203], + device='cuda:0'), covar=tensor([0.1125, 0.1552, 0.0365, 0.2212, 0.0886, 0.0309, 0.0692, 0.0551], + device='cuda:0'), in_proj_covar=tensor([0.0303, 0.0325, 0.0289, 0.0319, 0.0319, 0.0276, 0.0434, 0.0305], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 10:25:18,508 INFO [optim.py:369] (0/4) 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 10:25:28,313 INFO [train.py:901] (0/4) Epoch 26, batch 4650, loss[loss=0.2395, simple_loss=0.319, pruned_loss=0.07998, over 8661.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2842, pruned_loss=0.05883, over 1619454.33 frames. ], batch size: 49, lr: 2.90e-03, grad_scale: 8.0 +2023-02-07 10:26:02,070 INFO [train.py:901] (0/4) Epoch 26, batch 4700, loss[loss=0.2093, simple_loss=0.2888, pruned_loss=0.06493, over 7788.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2829, pruned_loss=0.0578, over 1619211.53 frames. ], batch size: 19, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:26:12,877 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0127, 1.7481, 2.9896, 1.6918, 2.4116, 3.2583, 3.3255, 2.8560], + device='cuda:0'), covar=tensor([0.1074, 0.1639, 0.0431, 0.1874, 0.1185, 0.0265, 0.0598, 0.0495], + device='cuda:0'), in_proj_covar=tensor([0.0303, 0.0326, 0.0289, 0.0319, 0.0319, 0.0276, 0.0435, 0.0306], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 10:26:13,614 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 10:26:32,494 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=206814.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:26:37,697 INFO [train.py:901] (0/4) Epoch 26, batch 4750, loss[loss=0.2038, simple_loss=0.2937, pruned_loss=0.05694, over 8513.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2834, pruned_loss=0.05839, over 1617867.37 frames. ], batch size: 49, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:26:53,338 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-07 10:26:55,372 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-07 10:26:58,821 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=206852.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:27:12,088 INFO [train.py:901] (0/4) Epoch 26, batch 4800, loss[loss=0.1879, simple_loss=0.2819, pruned_loss=0.047, over 8463.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.283, pruned_loss=0.05823, over 1618116.20 frames. ], batch size: 25, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:27:37,299 INFO [optim.py:369] (0/4) 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 10:27:46,704 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-07 10:27:47,383 INFO [train.py:901] (0/4) Epoch 26, batch 4850, loss[loss=0.1633, simple_loss=0.245, pruned_loss=0.04078, over 7808.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2826, pruned_loss=0.05839, over 1615821.35 frames. ], batch size: 20, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:27:52,976 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=206930.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:28:10,401 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=206955.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:28:21,661 INFO [train.py:901] (0/4) Epoch 26, batch 4900, loss[loss=0.2147, simple_loss=0.3002, pruned_loss=0.06455, over 8189.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2829, pruned_loss=0.05873, over 1613917.23 frames. ], batch size: 23, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:28:46,039 INFO [optim.py:369] (0/4) 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 10:28:46,439 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.49 vs. limit=2.0 +2023-02-07 10:28:55,267 INFO [train.py:901] (0/4) Epoch 26, batch 4950, loss[loss=0.1958, simple_loss=0.2758, pruned_loss=0.05792, over 8235.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.284, pruned_loss=0.05893, over 1616118.30 frames. ], batch size: 22, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:29:18,676 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.2400, 3.1575, 2.9746, 1.5528, 2.9173, 2.9178, 2.8146, 2.8026], + device='cuda:0'), covar=tensor([0.1224, 0.0892, 0.1371, 0.4536, 0.1169, 0.1339, 0.1861, 0.1066], + device='cuda:0'), in_proj_covar=tensor([0.0541, 0.0455, 0.0443, 0.0553, 0.0439, 0.0461, 0.0438, 0.0403], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 10:29:32,340 INFO [train.py:901] (0/4) Epoch 26, batch 5000, loss[loss=0.1608, simple_loss=0.2461, pruned_loss=0.0377, over 7805.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2853, pruned_loss=0.05983, over 1618702.30 frames. ], batch size: 20, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:29:57,378 INFO [optim.py:369] (0/4) 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 10:30:06,446 INFO [train.py:901] (0/4) Epoch 26, batch 5050, loss[loss=0.2192, simple_loss=0.3068, pruned_loss=0.06582, over 8335.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2853, pruned_loss=0.05981, over 1614871.25 frames. ], batch size: 26, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:30:24,789 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-07 10:30:42,637 INFO [train.py:901] (0/4) Epoch 26, batch 5100, loss[loss=0.2003, simple_loss=0.2855, pruned_loss=0.05754, over 8636.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.284, pruned_loss=0.0591, over 1613945.79 frames. ], batch size: 49, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:30:58,150 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=207194.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:30:58,765 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.8633, 6.0112, 5.2511, 2.5114, 5.3140, 5.5793, 5.4631, 5.3782], + device='cuda:0'), covar=tensor([0.0490, 0.0315, 0.0866, 0.4072, 0.0761, 0.0792, 0.1051, 0.0591], + device='cuda:0'), in_proj_covar=tensor([0.0538, 0.0453, 0.0440, 0.0550, 0.0437, 0.0457, 0.0434, 0.0400], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 10:30:59,431 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=207196.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:31:08,230 INFO [optim.py:369] (0/4) 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 10:31:16,935 INFO [train.py:901] (0/4) Epoch 26, batch 5150, loss[loss=0.1909, simple_loss=0.2816, pruned_loss=0.05007, over 8467.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2839, pruned_loss=0.05924, over 1613746.10 frames. ], batch size: 27, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:31:34,006 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5750, 1.8659, 1.8846, 1.2622, 1.9214, 1.4189, 0.4342, 1.7762], + device='cuda:0'), covar=tensor([0.0538, 0.0377, 0.0301, 0.0592, 0.0461, 0.1004, 0.0991, 0.0354], + device='cuda:0'), in_proj_covar=tensor([0.0468, 0.0407, 0.0361, 0.0459, 0.0392, 0.0551, 0.0403, 0.0437], + device='cuda:0'), out_proj_covar=tensor([1.2442e-04, 1.0592e-04, 9.4180e-05, 1.2007e-04, 1.0281e-04, 1.5430e-04, + 1.0792e-04, 1.1494e-04], device='cuda:0') +2023-02-07 10:31:52,790 INFO [train.py:901] (0/4) Epoch 26, batch 5200, loss[loss=0.2195, simple_loss=0.3105, pruned_loss=0.06421, over 8286.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2832, pruned_loss=0.05894, over 1612339.67 frames. ], batch size: 23, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:32:09,867 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 +2023-02-07 10:32:17,989 INFO [optim.py:369] (0/4) 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 10:32:19,447 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=207311.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:32:19,957 WARNING [train.py:1067] (0/4) Exclude cut with ID 7859-102521-0017-21930_sp0.9 from training. Duration: 27.25 +2023-02-07 10:32:26,697 INFO [train.py:901] (0/4) Epoch 26, batch 5250, loss[loss=0.2141, simple_loss=0.2943, pruned_loss=0.06693, over 8606.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2833, pruned_loss=0.05909, over 1613666.01 frames. ], batch size: 34, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:32:29,621 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5358, 2.4261, 1.7053, 2.1987, 2.0680, 1.5455, 1.9452, 2.0251], + device='cuda:0'), covar=tensor([0.1635, 0.0470, 0.1402, 0.0737, 0.0965, 0.1821, 0.1265, 0.1118], + device='cuda:0'), in_proj_covar=tensor([0.0361, 0.0241, 0.0340, 0.0312, 0.0302, 0.0347, 0.0349, 0.0324], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 10:32:50,482 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.9424, 1.6908, 6.0727, 2.1205, 5.4886, 5.1322, 5.6028, 5.4880], + device='cuda:0'), covar=tensor([0.0449, 0.4915, 0.0352, 0.4199, 0.0958, 0.0905, 0.0507, 0.0553], + device='cuda:0'), in_proj_covar=tensor([0.0665, 0.0661, 0.0730, 0.0652, 0.0739, 0.0628, 0.0629, 0.0703], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 10:33:00,346 INFO [train.py:901] (0/4) Epoch 26, batch 5300, loss[loss=0.2238, simple_loss=0.3084, pruned_loss=0.06954, over 8501.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2841, pruned_loss=0.05942, over 1614086.55 frames. ], batch size: 28, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:33:27,789 INFO [optim.py:369] (0/4) 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 10:33:36,843 INFO [train.py:901] (0/4) Epoch 26, batch 5350, loss[loss=0.1904, simple_loss=0.2739, pruned_loss=0.05341, over 8253.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2823, pruned_loss=0.05832, over 1606856.95 frames. ], batch size: 24, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:34:10,280 INFO [train.py:901] (0/4) Epoch 26, batch 5400, loss[loss=0.1745, simple_loss=0.2596, pruned_loss=0.04473, over 7792.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2815, pruned_loss=0.05726, over 1607892.30 frames. ], batch size: 19, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:34:37,323 INFO [optim.py:369] (0/4) 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 10:34:46,157 INFO [train.py:901] (0/4) Epoch 26, batch 5450, loss[loss=0.1776, simple_loss=0.2551, pruned_loss=0.04999, over 7444.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2823, pruned_loss=0.05755, over 1610286.67 frames. ], batch size: 17, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:34:56,175 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 +2023-02-07 10:34:57,839 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=207538.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:35:06,590 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-07 10:35:17,344 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=207567.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:35:20,580 INFO [train.py:901] (0/4) Epoch 26, batch 5500, loss[loss=0.1898, simple_loss=0.2822, pruned_loss=0.04865, over 8130.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2827, pruned_loss=0.05754, over 1614200.59 frames. ], batch size: 22, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:35:31,007 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8058, 2.0345, 2.1638, 1.4648, 2.2717, 1.5356, 0.7117, 2.0462], + device='cuda:0'), covar=tensor([0.0701, 0.0417, 0.0352, 0.0703, 0.0515, 0.1032, 0.1026, 0.0342], + device='cuda:0'), in_proj_covar=tensor([0.0464, 0.0403, 0.0358, 0.0455, 0.0389, 0.0546, 0.0400, 0.0432], + device='cuda:0'), out_proj_covar=tensor([1.2314e-04, 1.0483e-04, 9.3501e-05, 1.1919e-04, 1.0202e-04, 1.5287e-04, + 1.0721e-04, 1.1351e-04], device='cuda:0') +2023-02-07 10:35:34,396 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 10:35:56,709 INFO [train.py:901] (0/4) Epoch 26, batch 5550, loss[loss=0.1629, simple_loss=0.242, pruned_loss=0.04188, over 7701.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2829, pruned_loss=0.05711, over 1616929.02 frames. ], batch size: 18, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:36:17,870 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=207653.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:36:30,424 INFO [train.py:901] (0/4) Epoch 26, batch 5600, loss[loss=0.1841, simple_loss=0.2588, pruned_loss=0.0547, over 7251.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2834, pruned_loss=0.05736, over 1619389.89 frames. ], batch size: 16, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:36:55,074 INFO [optim.py:369] (0/4) 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 10:37:01,326 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3310, 2.1214, 1.7804, 2.0342, 1.6400, 1.5126, 1.6726, 1.7719], + device='cuda:0'), covar=tensor([0.1327, 0.0492, 0.1195, 0.0521, 0.0910, 0.1594, 0.0954, 0.0759], + device='cuda:0'), in_proj_covar=tensor([0.0360, 0.0243, 0.0341, 0.0313, 0.0303, 0.0347, 0.0349, 0.0325], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 10:37:04,589 INFO [train.py:901] (0/4) Epoch 26, batch 5650, loss[loss=0.2017, simple_loss=0.2783, pruned_loss=0.06262, over 7538.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2827, pruned_loss=0.05688, over 1616259.78 frames. ], batch size: 18, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:37:12,998 WARNING [train.py:1067] (0/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-07 10:37:40,858 INFO [train.py:901] (0/4) Epoch 26, batch 5700, loss[loss=0.2098, simple_loss=0.2973, pruned_loss=0.06116, over 8591.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2819, pruned_loss=0.05661, over 1615501.75 frames. ], batch size: 39, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:37:52,449 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3213, 1.6633, 1.7131, 0.9803, 1.6553, 1.3428, 0.2445, 1.5284], + device='cuda:0'), covar=tensor([0.0536, 0.0395, 0.0316, 0.0583, 0.0446, 0.0942, 0.0941, 0.0326], + device='cuda:0'), in_proj_covar=tensor([0.0465, 0.0402, 0.0359, 0.0456, 0.0389, 0.0547, 0.0401, 0.0432], + device='cuda:0'), out_proj_covar=tensor([1.2356e-04, 1.0471e-04, 9.3756e-05, 1.1947e-04, 1.0200e-04, 1.5306e-04, + 1.0733e-04, 1.1359e-04], device='cuda:0') +2023-02-07 10:38:05,994 INFO [optim.py:369] (0/4) 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 10:38:14,797 INFO [train.py:901] (0/4) Epoch 26, batch 5750, loss[loss=0.2078, simple_loss=0.2953, pruned_loss=0.06016, over 8517.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2818, pruned_loss=0.05703, over 1615494.96 frames. ], batch size: 26, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:38:16,870 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-07 10:38:50,530 INFO [train.py:901] (0/4) Epoch 26, batch 5800, loss[loss=0.1907, simple_loss=0.2899, pruned_loss=0.04579, over 8110.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2802, pruned_loss=0.05635, over 1611357.72 frames. ], batch size: 23, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:39:15,988 INFO [optim.py:369] (0/4) 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 10:39:16,254 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=207909.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:39:17,563 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8791, 1.7819, 2.5316, 1.6344, 1.4257, 2.5148, 0.5580, 1.5526], + device='cuda:0'), covar=tensor([0.1645, 0.1074, 0.0309, 0.1194, 0.2195, 0.0326, 0.1816, 0.1219], + device='cuda:0'), in_proj_covar=tensor([0.0198, 0.0204, 0.0133, 0.0223, 0.0277, 0.0145, 0.0171, 0.0198], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 10:39:24,803 INFO [train.py:901] (0/4) Epoch 26, batch 5850, loss[loss=0.1937, simple_loss=0.2927, pruned_loss=0.04738, over 8503.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2804, pruned_loss=0.05647, over 1613461.44 frames. ], batch size: 26, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:39:25,028 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8378, 2.1384, 2.1893, 1.4671, 2.2438, 1.7487, 0.6667, 1.9973], + device='cuda:0'), covar=tensor([0.0611, 0.0399, 0.0339, 0.0598, 0.0505, 0.0848, 0.0949, 0.0309], + device='cuda:0'), in_proj_covar=tensor([0.0467, 0.0403, 0.0361, 0.0458, 0.0391, 0.0549, 0.0403, 0.0435], + device='cuda:0'), out_proj_covar=tensor([1.2413e-04, 1.0500e-04, 9.4186e-05, 1.1985e-04, 1.0244e-04, 1.5341e-04, + 1.0788e-04, 1.1434e-04], device='cuda:0') +2023-02-07 10:39:32,992 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=207934.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:39:59,802 INFO [train.py:901] (0/4) Epoch 26, batch 5900, loss[loss=0.182, simple_loss=0.2739, pruned_loss=0.04502, over 8027.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2802, pruned_loss=0.05616, over 1615463.29 frames. ], batch size: 22, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:40:17,925 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8301, 2.1141, 2.1941, 1.5424, 2.2932, 1.6448, 0.8105, 1.9488], + device='cuda:0'), covar=tensor([0.0692, 0.0394, 0.0317, 0.0632, 0.0471, 0.0909, 0.0969, 0.0396], + device='cuda:0'), in_proj_covar=tensor([0.0468, 0.0403, 0.0360, 0.0456, 0.0390, 0.0547, 0.0402, 0.0434], + device='cuda:0'), out_proj_covar=tensor([1.2420e-04, 1.0499e-04, 9.4113e-05, 1.1951e-04, 1.0242e-04, 1.5293e-04, + 1.0772e-04, 1.1414e-04], device='cuda:0') +2023-02-07 10:40:19,157 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-208000.pt +2023-02-07 10:40:26,849 INFO [optim.py:369] (0/4) 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 10:40:35,639 INFO [train.py:901] (0/4) Epoch 26, batch 5950, loss[loss=0.1835, simple_loss=0.2536, pruned_loss=0.0567, over 7795.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2798, pruned_loss=0.05606, over 1614840.58 frames. ], batch size: 19, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:40:54,907 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=208050.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:41:01,456 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 +2023-02-07 10:41:07,267 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1209, 1.7652, 3.1673, 1.7416, 2.4095, 3.4805, 3.4989, 3.0444], + device='cuda:0'), covar=tensor([0.1070, 0.1581, 0.0378, 0.1863, 0.1129, 0.0221, 0.0576, 0.0462], + device='cuda:0'), in_proj_covar=tensor([0.0299, 0.0322, 0.0288, 0.0314, 0.0316, 0.0274, 0.0432, 0.0303], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 10:41:09,813 INFO [train.py:901] (0/4) Epoch 26, batch 6000, loss[loss=0.2135, simple_loss=0.3019, pruned_loss=0.06258, over 8506.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2808, pruned_loss=0.05693, over 1612499.91 frames. ], batch size: 26, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:41:09,814 INFO [train.py:926] (0/4) Computing validation loss +2023-02-07 10:41:24,452 INFO [train.py:935] (0/4) Epoch 26, validation: loss=0.1721, simple_loss=0.2717, pruned_loss=0.03627, over 944034.00 frames. +2023-02-07 10:41:24,453 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6760MB +2023-02-07 10:41:51,028 INFO [optim.py:369] (0/4) 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 10:42:00,869 INFO [train.py:901] (0/4) Epoch 26, batch 6050, loss[loss=0.1943, simple_loss=0.2891, pruned_loss=0.04976, over 8516.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.2814, pruned_loss=0.05675, over 1615902.76 frames. ], batch size: 26, lr: 2.89e-03, grad_scale: 8.0 +2023-02-07 10:42:25,307 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.9235, 2.3398, 3.6668, 1.9943, 1.8078, 3.5304, 0.6896, 2.1797], + device='cuda:0'), covar=tensor([0.1165, 0.1160, 0.0240, 0.1403, 0.2295, 0.0399, 0.1942, 0.1200], + device='cuda:0'), in_proj_covar=tensor([0.0196, 0.0204, 0.0133, 0.0221, 0.0275, 0.0144, 0.0170, 0.0197], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 10:42:27,456 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 +2023-02-07 10:42:36,027 INFO [train.py:901] (0/4) Epoch 26, batch 6100, loss[loss=0.207, simple_loss=0.2937, pruned_loss=0.06014, over 8320.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2813, pruned_loss=0.05681, over 1616050.57 frames. ], batch size: 25, lr: 2.88e-03, grad_scale: 8.0 +2023-02-07 10:42:48,575 WARNING [train.py:1067] (0/4) 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] (0/4) 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 10:43:05,095 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 +2023-02-07 10:43:10,769 INFO [train.py:901] (0/4) Epoch 26, batch 6150, loss[loss=0.1733, simple_loss=0.2624, pruned_loss=0.04207, over 8024.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2799, pruned_loss=0.05587, over 1614338.97 frames. ], batch size: 22, lr: 2.88e-03, grad_scale: 8.0 +2023-02-07 10:43:45,589 INFO [train.py:901] (0/4) Epoch 26, batch 6200, loss[loss=0.2248, simple_loss=0.3124, pruned_loss=0.06862, over 8657.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2808, pruned_loss=0.05652, over 1618230.32 frames. ], batch size: 34, lr: 2.88e-03, grad_scale: 8.0 +2023-02-07 10:43:53,058 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=208283.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:44:10,220 INFO [optim.py:369] (0/4) 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 10:44:19,004 INFO [train.py:901] (0/4) Epoch 26, batch 6250, loss[loss=0.2009, simple_loss=0.2824, pruned_loss=0.05967, over 8144.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2824, pruned_loss=0.05749, over 1618643.27 frames. ], batch size: 22, lr: 2.88e-03, grad_scale: 16.0 +2023-02-07 10:44:31,996 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.2394, 3.1604, 2.9281, 1.6816, 2.8736, 2.9504, 2.8808, 2.7743], + device='cuda:0'), covar=tensor([0.1252, 0.0887, 0.1396, 0.4337, 0.1159, 0.1296, 0.1654, 0.1047], + device='cuda:0'), in_proj_covar=tensor([0.0532, 0.0449, 0.0437, 0.0545, 0.0430, 0.0454, 0.0429, 0.0397], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 10:44:55,648 INFO [train.py:901] (0/4) Epoch 26, batch 6300, loss[loss=0.1999, simple_loss=0.2719, pruned_loss=0.06394, over 7644.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2821, pruned_loss=0.05791, over 1612687.85 frames. ], batch size: 19, lr: 2.88e-03, grad_scale: 16.0 +2023-02-07 10:45:10,443 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 10:45:25,955 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.8150, 1.4188, 3.9689, 1.4418, 3.5406, 3.2812, 3.6331, 3.5097], + device='cuda:0'), covar=tensor([0.0745, 0.4847, 0.0695, 0.4422, 0.1224, 0.1082, 0.0687, 0.0797], + device='cuda:0'), in_proj_covar=tensor([0.0665, 0.0660, 0.0728, 0.0652, 0.0737, 0.0629, 0.0627, 0.0706], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 10:45:29,139 INFO [train.py:901] (0/4) Epoch 26, batch 6350, loss[loss=0.1793, simple_loss=0.2556, pruned_loss=0.05147, over 7278.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2812, pruned_loss=0.05721, over 1613222.65 frames. ], batch size: 16, lr: 2.88e-03, grad_scale: 16.0 +2023-02-07 10:45:54,272 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6971, 1.4789, 3.0913, 1.3783, 2.2281, 3.3660, 3.5098, 2.8271], + device='cuda:0'), covar=tensor([0.1335, 0.1949, 0.0374, 0.2289, 0.1055, 0.0291, 0.0646, 0.0645], + device='cuda:0'), in_proj_covar=tensor([0.0302, 0.0324, 0.0289, 0.0317, 0.0318, 0.0276, 0.0434, 0.0305], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 10:46:04,931 INFO [train.py:901] (0/4) Epoch 26, batch 6400, loss[loss=0.2059, simple_loss=0.3017, pruned_loss=0.05505, over 8291.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2821, pruned_loss=0.0575, over 1616098.95 frames. ], batch size: 23, lr: 2.88e-03, grad_scale: 16.0 +2023-02-07 10:46:30,800 INFO [optim.py:369] (0/4) 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 10:46:30,997 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 26, batch 6450, loss[loss=0.1858, simple_loss=0.2701, pruned_loss=0.05075, over 7925.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2825, pruned_loss=0.05804, over 1610450.79 frames. ], batch size: 20, lr: 2.88e-03, grad_scale: 16.0 +2023-02-07 10:47:09,470 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6075, 2.6079, 1.8001, 2.3027, 2.2386, 1.6815, 2.0556, 2.1604], + device='cuda:0'), covar=tensor([0.1444, 0.0398, 0.1186, 0.0622, 0.0691, 0.1404, 0.1002, 0.1028], + device='cuda:0'), in_proj_covar=tensor([0.0359, 0.0241, 0.0340, 0.0312, 0.0303, 0.0347, 0.0349, 0.0324], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 10:47:11,953 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1942, 1.4657, 4.2578, 1.9171, 2.4827, 4.8682, 4.9730, 4.1410], + device='cuda:0'), covar=tensor([0.1212, 0.2109, 0.0310, 0.2104, 0.1248, 0.0199, 0.0498, 0.0598], + device='cuda:0'), in_proj_covar=tensor([0.0303, 0.0325, 0.0290, 0.0318, 0.0319, 0.0277, 0.0435, 0.0306], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 10:47:13,740 INFO [train.py:901] (0/4) Epoch 26, batch 6500, loss[loss=0.1927, simple_loss=0.2732, pruned_loss=0.05605, over 7929.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2826, pruned_loss=0.05788, over 1610550.19 frames. ], batch size: 20, lr: 2.88e-03, grad_scale: 16.0 +2023-02-07 10:47:27,387 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0277, 1.5516, 3.0948, 1.4060, 2.2819, 3.3410, 3.5050, 2.8445], + device='cuda:0'), covar=tensor([0.1093, 0.1819, 0.0422, 0.2286, 0.1034, 0.0318, 0.0672, 0.0577], + device='cuda:0'), in_proj_covar=tensor([0.0303, 0.0325, 0.0290, 0.0318, 0.0319, 0.0277, 0.0435, 0.0306], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 10:47:29,451 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=208594.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:47:39,224 INFO [optim.py:369] (0/4) 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 10:47:48,760 INFO [train.py:901] (0/4) Epoch 26, batch 6550, loss[loss=0.2041, simple_loss=0.297, pruned_loss=0.05553, over 8344.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2808, pruned_loss=0.05693, over 1609762.84 frames. ], batch size: 26, lr: 2.88e-03, grad_scale: 16.0 +2023-02-07 10:47:52,190 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=208627.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:47:56,162 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-07 10:48:12,571 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=2.05 vs. limit=2.0 +2023-02-07 10:48:12,889 WARNING [train.py:1067] (0/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-07 10:48:22,502 INFO [train.py:901] (0/4) Epoch 26, batch 6600, loss[loss=0.1917, simple_loss=0.2621, pruned_loss=0.06065, over 7974.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2815, pruned_loss=0.05752, over 1611837.24 frames. ], batch size: 21, lr: 2.88e-03, grad_scale: 16.0 +2023-02-07 10:48:44,618 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1103, 1.5239, 3.3337, 1.4658, 2.3144, 3.7012, 3.8040, 3.1330], + device='cuda:0'), covar=tensor([0.1186, 0.1917, 0.0350, 0.2240, 0.1074, 0.0263, 0.0608, 0.0582], + device='cuda:0'), in_proj_covar=tensor([0.0302, 0.0325, 0.0290, 0.0318, 0.0319, 0.0277, 0.0435, 0.0305], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 10:48:49,204 INFO [optim.py:369] (0/4) 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 10:48:58,721 INFO [train.py:901] (0/4) Epoch 26, batch 6650, loss[loss=0.1871, simple_loss=0.2645, pruned_loss=0.05487, over 7803.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2822, pruned_loss=0.0577, over 1618532.33 frames. ], batch size: 19, lr: 2.88e-03, grad_scale: 16.0 +2023-02-07 10:49:12,755 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=208742.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:49:28,859 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=208765.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:49:29,732 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.52 vs. limit=2.0 +2023-02-07 10:49:33,527 INFO [train.py:901] (0/4) Epoch 26, batch 6700, loss[loss=0.1951, simple_loss=0.2779, pruned_loss=0.05611, over 8101.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2823, pruned_loss=0.0576, over 1618868.07 frames. ], batch size: 23, lr: 2.88e-03, grad_scale: 16.0 +2023-02-07 10:49:46,225 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=208790.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:49:57,534 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 +2023-02-07 10:49:59,641 INFO [optim.py:369] (0/4) 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 10:50:09,371 INFO [train.py:901] (0/4) Epoch 26, batch 6750, loss[loss=0.1942, simple_loss=0.2593, pruned_loss=0.06451, over 7272.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2826, pruned_loss=0.05783, over 1615539.24 frames. ], batch size: 16, lr: 2.88e-03, grad_scale: 16.0 +2023-02-07 10:50:30,915 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-07 10:50:43,649 INFO [train.py:901] (0/4) Epoch 26, batch 6800, loss[loss=0.1919, simple_loss=0.2745, pruned_loss=0.05462, over 8089.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2826, pruned_loss=0.05776, over 1616871.13 frames. ], batch size: 21, lr: 2.88e-03, grad_scale: 16.0 +2023-02-07 10:51:05,253 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 +2023-02-07 10:51:08,812 INFO [optim.py:369] (0/4) 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 10:51:18,806 INFO [train.py:901] (0/4) Epoch 26, batch 6850, loss[loss=0.2364, simple_loss=0.314, pruned_loss=0.07935, over 8655.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.283, pruned_loss=0.05806, over 1616793.49 frames. ], batch size: 34, lr: 2.88e-03, grad_scale: 16.0 +2023-02-07 10:51:19,509 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-07 10:51:30,537 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=208938.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:51:43,005 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8319, 2.0541, 2.1183, 1.4305, 2.2575, 1.5821, 0.7731, 1.9561], + device='cuda:0'), covar=tensor([0.0675, 0.0418, 0.0357, 0.0717, 0.0451, 0.0952, 0.1062, 0.0383], + device='cuda:0'), in_proj_covar=tensor([0.0465, 0.0404, 0.0360, 0.0456, 0.0390, 0.0547, 0.0401, 0.0435], + device='cuda:0'), out_proj_covar=tensor([1.2348e-04, 1.0515e-04, 9.4029e-05, 1.1954e-04, 1.0199e-04, 1.5290e-04, + 1.0723e-04, 1.1427e-04], device='cuda:0') +2023-02-07 10:51:54,555 INFO [train.py:901] (0/4) Epoch 26, batch 6900, loss[loss=0.2257, simple_loss=0.2969, pruned_loss=0.07724, over 8339.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2835, pruned_loss=0.05811, over 1617931.33 frames. ], batch size: 26, lr: 2.88e-03, grad_scale: 8.0 +2023-02-07 10:52:12,013 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=208998.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:52:20,219 INFO [optim.py:369] (0/4) 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 10:52:23,020 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-07 10:52:28,982 INFO [train.py:901] (0/4) Epoch 26, batch 6950, loss[loss=0.2137, simple_loss=0.2968, pruned_loss=0.06528, over 8605.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2833, pruned_loss=0.05783, over 1619154.39 frames. ], batch size: 31, lr: 2.88e-03, grad_scale: 8.0 +2023-02-07 10:52:29,840 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209023.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:52:47,736 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9341, 1.7108, 2.8026, 2.1395, 2.6378, 1.9150, 1.7172, 1.4007], + device='cuda:0'), covar=tensor([0.7595, 0.6571, 0.2184, 0.4477, 0.3228, 0.4767, 0.3186, 0.6241], + device='cuda:0'), in_proj_covar=tensor([0.0960, 0.1013, 0.0829, 0.0985, 0.1016, 0.0922, 0.0769, 0.0846], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 10:52:51,120 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=209053.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:53:04,217 INFO [train.py:901] (0/4) Epoch 26, batch 7000, loss[loss=0.2297, simple_loss=0.3035, pruned_loss=0.07798, over 8029.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2822, pruned_loss=0.05781, over 1617155.41 frames. ], batch size: 22, lr: 2.88e-03, grad_scale: 8.0 +2023-02-07 10:53:07,235 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7441, 1.9886, 2.0906, 1.3991, 2.2051, 1.5573, 0.6514, 1.8638], + device='cuda:0'), covar=tensor([0.0636, 0.0409, 0.0362, 0.0680, 0.0424, 0.0957, 0.0976, 0.0382], + device='cuda:0'), in_proj_covar=tensor([0.0469, 0.0406, 0.0362, 0.0459, 0.0392, 0.0550, 0.0403, 0.0437], + device='cuda:0'), out_proj_covar=tensor([1.2439e-04, 1.0566e-04, 9.4474e-05, 1.2022e-04, 1.0275e-04, 1.5369e-04, + 1.0770e-04, 1.1488e-04], device='cuda:0') +2023-02-07 10:53:15,631 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 +2023-02-07 10:53:30,452 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 26, batch 7050, loss[loss=0.1728, simple_loss=0.2621, pruned_loss=0.04171, over 7678.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.283, pruned_loss=0.05816, over 1613964.89 frames. ], batch size: 18, lr: 2.88e-03, grad_scale: 8.0 +2023-02-07 10:54:15,173 INFO [train.py:901] (0/4) Epoch 26, batch 7100, loss[loss=0.2115, simple_loss=0.3043, pruned_loss=0.05934, over 8467.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2833, pruned_loss=0.05804, over 1611944.67 frames. ], batch size: 29, lr: 2.88e-03, grad_scale: 8.0 +2023-02-07 10:54:34,737 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8273, 2.0964, 2.1875, 1.4021, 2.3024, 1.5706, 0.7209, 1.9326], + device='cuda:0'), covar=tensor([0.0700, 0.0415, 0.0323, 0.0709, 0.0508, 0.0984, 0.0997, 0.0377], + device='cuda:0'), in_proj_covar=tensor([0.0468, 0.0404, 0.0360, 0.0457, 0.0390, 0.0548, 0.0402, 0.0436], + device='cuda:0'), out_proj_covar=tensor([1.2422e-04, 1.0530e-04, 9.4032e-05, 1.1962e-04, 1.0213e-04, 1.5309e-04, + 1.0742e-04, 1.1443e-04], device='cuda:0') +2023-02-07 10:54:42,204 INFO [optim.py:369] (0/4) 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 10:54:49,199 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209220.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:54:50,434 INFO [train.py:901] (0/4) Epoch 26, batch 7150, loss[loss=0.2065, simple_loss=0.2886, pruned_loss=0.06217, over 8531.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2826, pruned_loss=0.05767, over 1617144.88 frames. ], batch size: 28, lr: 2.88e-03, grad_scale: 8.0 +2023-02-07 10:54:56,930 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 +2023-02-07 10:55:03,374 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7443, 1.9707, 2.0843, 1.4147, 2.0813, 1.6368, 0.5756, 1.9475], + device='cuda:0'), covar=tensor([0.0565, 0.0353, 0.0302, 0.0579, 0.0467, 0.0903, 0.0911, 0.0269], + device='cuda:0'), in_proj_covar=tensor([0.0467, 0.0405, 0.0360, 0.0456, 0.0390, 0.0548, 0.0402, 0.0436], + device='cuda:0'), out_proj_covar=tensor([1.2414e-04, 1.0539e-04, 9.3979e-05, 1.1942e-04, 1.0217e-04, 1.5325e-04, + 1.0739e-04, 1.1444e-04], device='cuda:0') +2023-02-07 10:55:07,565 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.48 vs. limit=2.0 +2023-02-07 10:55:19,351 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6822, 2.2985, 3.8199, 1.5751, 2.8812, 2.1887, 1.9156, 2.7618], + device='cuda:0'), covar=tensor([0.1984, 0.2579, 0.0841, 0.4676, 0.1919, 0.3297, 0.2323, 0.2458], + device='cuda:0'), in_proj_covar=tensor([0.0535, 0.0630, 0.0560, 0.0665, 0.0659, 0.0609, 0.0558, 0.0645], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 10:55:20,030 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7625, 2.4151, 3.8676, 1.5442, 2.8550, 2.2300, 2.0080, 2.6792], + device='cuda:0'), covar=tensor([0.2056, 0.2735, 0.1003, 0.5140, 0.2054, 0.3537, 0.2594, 0.3021], + device='cuda:0'), in_proj_covar=tensor([0.0535, 0.0630, 0.0560, 0.0665, 0.0659, 0.0609, 0.0558, 0.0645], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 10:55:25,278 INFO [train.py:901] (0/4) Epoch 26, batch 7200, loss[loss=0.1687, simple_loss=0.2704, pruned_loss=0.03351, over 8733.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2837, pruned_loss=0.05813, over 1622998.45 frames. ], batch size: 39, lr: 2.88e-03, grad_scale: 8.0 +2023-02-07 10:55:25,475 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([0.9978, 1.6287, 1.4454, 1.5606, 1.3479, 1.3164, 1.3550, 1.2556], + device='cuda:0'), covar=tensor([0.1315, 0.0554, 0.1357, 0.0599, 0.0823, 0.1628, 0.0974, 0.0887], + device='cuda:0'), in_proj_covar=tensor([0.0360, 0.0239, 0.0340, 0.0312, 0.0302, 0.0345, 0.0347, 0.0323], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 10:55:26,114 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209273.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:55:26,746 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.8461, 5.9978, 5.1932, 2.3382, 5.2053, 5.5676, 5.4991, 5.3889], + device='cuda:0'), covar=tensor([0.0495, 0.0386, 0.0786, 0.4301, 0.0673, 0.0766, 0.0970, 0.0547], + device='cuda:0'), in_proj_covar=tensor([0.0538, 0.0457, 0.0443, 0.0554, 0.0437, 0.0461, 0.0436, 0.0403], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 10:55:51,810 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 10:55:56,506 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 26, batch 7250, loss[loss=0.263, simple_loss=0.3346, pruned_loss=0.09573, over 8026.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2841, pruned_loss=0.05848, over 1622149.58 frames. ], batch size: 22, lr: 2.88e-03, grad_scale: 8.0 +2023-02-07 10:56:08,436 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 26, batch 7300, loss[loss=0.1799, simple_loss=0.2667, pruned_loss=0.04654, over 8248.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2836, pruned_loss=0.05803, over 1623087.53 frames. ], batch size: 22, lr: 2.88e-03, grad_scale: 8.0 +2023-02-07 10:56:59,845 INFO [optim.py:369] (0/4) 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 10:57:06,766 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-07 10:57:09,434 INFO [train.py:901] (0/4) Epoch 26, batch 7350, loss[loss=0.1892, simple_loss=0.2656, pruned_loss=0.05646, over 7657.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2853, pruned_loss=0.05943, over 1619939.86 frames. ], batch size: 19, lr: 2.88e-03, grad_scale: 8.0 +2023-02-07 10:57:26,313 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 26, batch 7400, loss[loss=0.1895, simple_loss=0.2749, pruned_loss=0.05203, over 7966.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2849, pruned_loss=0.05931, over 1621863.59 frames. ], batch size: 21, lr: 2.88e-03, grad_scale: 8.0 +2023-02-07 10:57:50,999 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209484.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:58:04,816 WARNING [train.py:1067] (0/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-07 10:58:09,480 INFO [optim.py:369] (0/4) 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 10:58:17,540 INFO [train.py:901] (0/4) Epoch 26, batch 7450, loss[loss=0.2385, simple_loss=0.3218, pruned_loss=0.07761, over 8608.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2854, pruned_loss=0.05973, over 1622964.71 frames. ], batch size: 31, lr: 2.88e-03, grad_scale: 8.0 +2023-02-07 10:58:22,475 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209528.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:58:47,395 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=209564.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:58:52,742 INFO [train.py:901] (0/4) Epoch 26, batch 7500, loss[loss=0.1934, simple_loss=0.2761, pruned_loss=0.05535, over 8186.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2849, pruned_loss=0.05984, over 1618574.41 frames. ], batch size: 23, lr: 2.88e-03, grad_scale: 8.0 +2023-02-07 10:59:18,692 INFO [optim.py:369] (0/4) 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 10:59:24,163 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 26, batch 7550, loss[loss=0.1882, simple_loss=0.2689, pruned_loss=0.05374, over 7935.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2848, pruned_loss=0.05974, over 1617808.26 frames. ], batch size: 20, lr: 2.87e-03, grad_scale: 8.0 +2023-02-07 10:59:28,776 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209624.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:59:34,807 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 +2023-02-07 10:59:41,889 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5121, 2.4515, 1.7399, 2.2426, 1.9987, 1.4277, 1.9684, 2.1955], + device='cuda:0'), covar=tensor([0.1964, 0.0586, 0.1555, 0.0753, 0.1083, 0.2044, 0.1370, 0.1123], + device='cuda:0'), in_proj_covar=tensor([0.0363, 0.0241, 0.0342, 0.0314, 0.0306, 0.0349, 0.0351, 0.0326], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 10:59:54,103 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=209660.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 10:59:55,166 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 +2023-02-07 11:00:01,942 INFO [train.py:901] (0/4) Epoch 26, batch 7600, loss[loss=0.1958, simple_loss=0.2845, pruned_loss=0.05352, over 8109.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2853, pruned_loss=0.05971, over 1615807.33 frames. ], batch size: 23, lr: 2.87e-03, grad_scale: 8.0 +2023-02-07 11:00:07,012 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=209679.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:00:25,092 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209706.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:00:27,672 INFO [optim.py:369] (0/4) 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 11:00:35,725 INFO [train.py:901] (0/4) Epoch 26, batch 7650, loss[loss=0.2087, simple_loss=0.2892, pruned_loss=0.06408, over 8030.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2856, pruned_loss=0.05978, over 1622518.64 frames. ], batch size: 22, lr: 2.87e-03, grad_scale: 8.0 +2023-02-07 11:00:43,054 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=209732.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:00:52,377 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.5691, 1.4675, 4.7432, 1.8384, 4.2806, 3.9278, 4.2898, 4.2305], + device='cuda:0'), covar=tensor([0.0597, 0.4990, 0.0546, 0.4172, 0.0995, 0.1049, 0.0630, 0.0655], + device='cuda:0'), in_proj_covar=tensor([0.0674, 0.0665, 0.0736, 0.0659, 0.0743, 0.0635, 0.0635, 0.0711], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 11:01:10,731 INFO [train.py:901] (0/4) Epoch 26, batch 7700, loss[loss=0.1671, simple_loss=0.2475, pruned_loss=0.0434, over 7650.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2861, pruned_loss=0.05999, over 1619213.44 frames. ], batch size: 19, lr: 2.87e-03, grad_scale: 8.0 +2023-02-07 11:01:12,816 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.6470, 4.6670, 4.2184, 2.1746, 4.1431, 4.2593, 4.1866, 4.0374], + device='cuda:0'), covar=tensor([0.0635, 0.0493, 0.1099, 0.4098, 0.0772, 0.1073, 0.1250, 0.0681], + device='cuda:0'), in_proj_covar=tensor([0.0538, 0.0454, 0.0442, 0.0553, 0.0435, 0.0460, 0.0435, 0.0401], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 11:01:12,896 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=209775.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:01:14,129 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0009-44385_sp0.9 from training. Duration: 27.02225 +2023-02-07 11:01:36,864 INFO [optim.py:369] (0/4) 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 11:01:44,893 INFO [train.py:901] (0/4) Epoch 26, batch 7750, loss[loss=0.1931, simple_loss=0.2714, pruned_loss=0.05743, over 8463.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2854, pruned_loss=0.05997, over 1615132.87 frames. ], batch size: 25, lr: 2.87e-03, grad_scale: 8.0 +2023-02-07 11:01:48,925 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=209828.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:02:19,758 INFO [train.py:901] (0/4) Epoch 26, batch 7800, loss[loss=0.1987, simple_loss=0.2854, pruned_loss=0.05599, over 7943.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2846, pruned_loss=0.0595, over 1614209.10 frames. ], batch size: 20, lr: 2.87e-03, grad_scale: 8.0 +2023-02-07 11:02:19,819 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=209872.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:02:45,322 INFO [optim.py:369] (0/4) 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 11:02:53,247 INFO [train.py:901] (0/4) Epoch 26, batch 7850, loss[loss=0.1659, simple_loss=0.2495, pruned_loss=0.04117, over 7935.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2833, pruned_loss=0.05877, over 1614913.78 frames. ], batch size: 20, lr: 2.87e-03, grad_scale: 8.0 +2023-02-07 11:03:01,890 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209935.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:03:07,202 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=209943.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:03:18,173 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209960.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:03:23,276 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=209968.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:03:25,899 INFO [train.py:901] (0/4) Epoch 26, batch 7900, loss[loss=0.1643, simple_loss=0.2389, pruned_loss=0.04488, over 7805.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2819, pruned_loss=0.05813, over 1609383.40 frames. ], batch size: 19, lr: 2.87e-03, grad_scale: 8.0 +2023-02-07 11:03:35,832 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=209987.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:03:36,567 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209988.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:03:37,798 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.2269, 3.1398, 2.9697, 1.4540, 2.8731, 2.9839, 2.8908, 2.8247], + device='cuda:0'), covar=tensor([0.1250, 0.0877, 0.1399, 0.4982, 0.1065, 0.1162, 0.1650, 0.0980], + device='cuda:0'), in_proj_covar=tensor([0.0533, 0.0452, 0.0438, 0.0548, 0.0433, 0.0456, 0.0431, 0.0398], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 11:03:44,410 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-210000.pt +2023-02-07 11:03:51,912 INFO [optim.py:369] (0/4) 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 11:03:54,071 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210013.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:03:59,748 INFO [train.py:901] (0/4) Epoch 26, batch 7950, loss[loss=0.2042, simple_loss=0.2916, pruned_loss=0.05845, over 8192.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2826, pruned_loss=0.05849, over 1611241.38 frames. ], batch size: 23, lr: 2.87e-03, grad_scale: 8.0 +2023-02-07 11:04:06,000 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=210031.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:04:18,574 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=210050.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:04:22,762 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210056.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:04:26,359 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.80 vs. limit=5.0 +2023-02-07 11:04:33,085 INFO [train.py:901] (0/4) Epoch 26, batch 8000, loss[loss=0.1825, simple_loss=0.2765, pruned_loss=0.04426, over 8288.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2822, pruned_loss=0.05791, over 1610779.94 frames. ], batch size: 23, lr: 2.87e-03, grad_scale: 8.0 +2023-02-07 11:04:35,221 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=210075.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:04:40,639 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=210083.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:04:51,786 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.21 vs. limit=5.0 +2023-02-07 11:04:53,029 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6598, 2.0856, 3.0729, 1.5630, 2.4594, 2.1462, 1.8356, 2.4087], + device='cuda:0'), covar=tensor([0.1947, 0.2560, 0.1005, 0.4626, 0.1789, 0.3171, 0.2349, 0.2204], + device='cuda:0'), in_proj_covar=tensor([0.0538, 0.0634, 0.0562, 0.0669, 0.0662, 0.0612, 0.0561, 0.0645], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 11:04:58,003 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 26, batch 8050, loss[loss=0.1923, simple_loss=0.2673, pruned_loss=0.05868, over 7550.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.2793, pruned_loss=0.05692, over 1592142.18 frames. ], batch size: 18, lr: 2.87e-03, grad_scale: 8.0 +2023-02-07 11:05:21,063 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 +2023-02-07 11:05:21,922 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0783, 1.5341, 1.7331, 1.4442, 0.9970, 1.4852, 1.8773, 1.6520], + device='cuda:0'), covar=tensor([0.0534, 0.1281, 0.1723, 0.1462, 0.0594, 0.1515, 0.0669, 0.0667], + device='cuda:0'), in_proj_covar=tensor([0.0099, 0.0153, 0.0189, 0.0162, 0.0101, 0.0164, 0.0113, 0.0146], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:0') +2023-02-07 11:05:28,104 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/epoch-26.pt +2023-02-07 11:05:39,502 WARNING [train.py:1067] (0/4) Exclude cut with ID 3488-85273-0017-111273_sp0.9 from training. Duration: 27.47775 +2023-02-07 11:05:43,126 INFO [train.py:901] (0/4) Epoch 27, batch 0, loss[loss=0.1873, simple_loss=0.2663, pruned_loss=0.05414, over 7928.00 frames. ], tot_loss[loss=0.1873, simple_loss=0.2663, pruned_loss=0.05414, over 7928.00 frames. ], batch size: 20, lr: 2.82e-03, grad_scale: 8.0 +2023-02-07 11:05:43,127 INFO [train.py:926] (0/4) Computing validation loss +2023-02-07 11:05:54,197 INFO [train.py:935] (0/4) Epoch 27, validation: loss=0.172, simple_loss=0.2713, pruned_loss=0.03628, over 944034.00 frames. +2023-02-07 11:05:54,198 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6760MB +2023-02-07 11:06:01,164 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=210165.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:06:07,185 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4177, 1.2300, 2.3749, 1.2984, 2.2253, 2.5760, 2.7089, 2.1332], + device='cuda:0'), covar=tensor([0.1154, 0.1471, 0.0412, 0.2112, 0.0694, 0.0373, 0.0624, 0.0676], + device='cuda:0'), in_proj_covar=tensor([0.0303, 0.0325, 0.0290, 0.0319, 0.0317, 0.0276, 0.0433, 0.0305], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 11:06:08,365 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590_sp0.9 from training. Duration: 28.72225 +2023-02-07 11:06:24,779 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=210199.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:06:28,659 INFO [train.py:901] (0/4) Epoch 27, batch 50, loss[loss=0.2467, simple_loss=0.3348, pruned_loss=0.07936, over 8604.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2876, pruned_loss=0.0628, over 366206.96 frames. ], batch size: 31, lr: 2.82e-03, grad_scale: 8.0 +2023-02-07 11:06:33,573 INFO [optim.py:369] (0/4) 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] (0/4) Exclude cut with ID 6709-74022-0004-57021_sp1.1 from training. Duration: 0.9409375 +2023-02-07 11:06:43,317 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210224.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:06:56,860 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=210243.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:06:59,547 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1365, 1.8959, 2.3428, 2.0438, 2.3614, 2.2082, 2.0313, 1.1441], + device='cuda:0'), covar=tensor([0.5473, 0.4470, 0.2098, 0.3569, 0.2400, 0.3060, 0.1861, 0.4837], + device='cuda:0'), in_proj_covar=tensor([0.0962, 0.1018, 0.0828, 0.0987, 0.1019, 0.0926, 0.0772, 0.0847], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 11:07:04,531 INFO [train.py:901] (0/4) Epoch 27, batch 100, loss[loss=0.1989, simple_loss=0.2817, pruned_loss=0.05809, over 8137.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2863, pruned_loss=0.06106, over 644343.43 frames. ], batch size: 22, lr: 2.82e-03, grad_scale: 8.0 +2023-02-07 11:07:04,986 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 +2023-02-07 11:07:05,174 WARNING [train.py:1067] (0/4) Exclude cut with ID 497-129325-0061-9566_sp1.1 from training. Duration: 0.97725 +2023-02-07 11:07:13,374 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210268.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:07:22,921 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 +2023-02-07 11:07:34,280 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5981, 2.0758, 3.2751, 1.5374, 2.5274, 2.1819, 1.7463, 2.6277], + device='cuda:0'), covar=tensor([0.2128, 0.2840, 0.1019, 0.4702, 0.2018, 0.3348, 0.2584, 0.2297], + device='cuda:0'), in_proj_covar=tensor([0.0536, 0.0633, 0.0562, 0.0666, 0.0661, 0.0610, 0.0560, 0.0644], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 11:07:37,998 INFO [train.py:901] (0/4) Epoch 27, batch 150, loss[loss=0.1955, simple_loss=0.283, pruned_loss=0.05401, over 8346.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2837, pruned_loss=0.05884, over 861249.53 frames. ], batch size: 26, lr: 2.82e-03, grad_scale: 8.0 +2023-02-07 11:07:40,614 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.5620, 4.5519, 4.1285, 2.0567, 3.9702, 4.2195, 4.1209, 4.0392], + device='cuda:0'), covar=tensor([0.0662, 0.0530, 0.1003, 0.4448, 0.0848, 0.0900, 0.1191, 0.0717], + device='cuda:0'), in_proj_covar=tensor([0.0535, 0.0454, 0.0439, 0.0549, 0.0433, 0.0456, 0.0432, 0.0400], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 11:07:40,870 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 +2023-02-07 11:07:41,159 INFO [optim.py:369] (0/4) 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 11:07:46,929 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-02-07 11:08:02,853 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 27, batch 200, loss[loss=0.1842, simple_loss=0.2665, pruned_loss=0.05093, over 8088.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2853, pruned_loss=0.05919, over 1032228.89 frames. ], batch size: 21, lr: 2.82e-03, grad_scale: 8.0 +2023-02-07 11:08:20,475 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210364.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:08:48,339 INFO [train.py:901] (0/4) Epoch 27, batch 250, loss[loss=0.203, simple_loss=0.2946, pruned_loss=0.05567, over 8459.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.284, pruned_loss=0.05815, over 1167114.04 frames. ], batch size: 25, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:08:51,582 INFO [optim.py:369] (0/4) 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] (0/4) Exclude cut with ID 6951-79737-0043-83149_sp0.9 from training. Duration: 28.0944375 +2023-02-07 11:08:57,626 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=210419.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:08:59,057 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=210421.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:09:06,461 WARNING [train.py:1067] (0/4) Exclude cut with ID 4278-13270-0009-62705_sp0.9 from training. Duration: 25.45 +2023-02-07 11:09:15,915 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210446.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:09:23,005 INFO [train.py:901] (0/4) Epoch 27, batch 300, loss[loss=0.2034, simple_loss=0.2723, pruned_loss=0.06729, over 6747.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2842, pruned_loss=0.05885, over 1263805.73 frames. ], batch size: 15, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:09:26,773 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 +2023-02-07 11:09:46,152 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0593, 1.6015, 1.4884, 1.5893, 1.2759, 1.3327, 1.3123, 1.3635], + device='cuda:0'), covar=tensor([0.1135, 0.0468, 0.1238, 0.0547, 0.0799, 0.1451, 0.0911, 0.0723], + device='cuda:0'), in_proj_covar=tensor([0.0361, 0.0240, 0.0341, 0.0311, 0.0304, 0.0346, 0.0348, 0.0322], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 11:09:46,850 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7318, 2.6450, 2.0228, 2.3565, 2.1806, 1.7820, 2.1905, 2.2610], + device='cuda:0'), covar=tensor([0.1307, 0.0388, 0.1055, 0.0594, 0.0746, 0.1400, 0.0896, 0.0790], + device='cuda:0'), in_proj_covar=tensor([0.0361, 0.0240, 0.0341, 0.0311, 0.0304, 0.0346, 0.0348, 0.0322], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 11:09:57,348 INFO [train.py:901] (0/4) Epoch 27, batch 350, loss[loss=0.1939, simple_loss=0.2774, pruned_loss=0.05516, over 8464.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2833, pruned_loss=0.05863, over 1334975.27 frames. ], batch size: 25, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:09:57,505 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8734, 1.4384, 3.1097, 1.4046, 2.3206, 3.3771, 3.5025, 2.8893], + device='cuda:0'), covar=tensor([0.1209, 0.1911, 0.0373, 0.2207, 0.1039, 0.0260, 0.0590, 0.0526], + device='cuda:0'), in_proj_covar=tensor([0.0302, 0.0323, 0.0288, 0.0317, 0.0316, 0.0274, 0.0431, 0.0304], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 11:10:00,693 INFO [optim.py:369] (0/4) 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 11:10:01,074 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-07 11:10:08,251 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0837, 2.2703, 1.8260, 2.9139, 1.4617, 1.7344, 2.2125, 2.2672], + device='cuda:0'), covar=tensor([0.0752, 0.0795, 0.0923, 0.0360, 0.1108, 0.1315, 0.0845, 0.0853], + device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0194, 0.0246, 0.0211, 0.0202, 0.0245, 0.0249, 0.0205], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 11:10:16,892 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=210534.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:10:30,766 INFO [train.py:901] (0/4) Epoch 27, batch 400, loss[loss=0.238, simple_loss=0.3231, pruned_loss=0.07646, over 8487.00 frames. ], tot_loss[loss=0.202, simple_loss=0.285, pruned_loss=0.05947, over 1396158.96 frames. ], batch size: 29, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:11:06,883 INFO [train.py:901] (0/4) Epoch 27, batch 450, loss[loss=0.1482, simple_loss=0.2281, pruned_loss=0.03411, over 7696.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2839, pruned_loss=0.0587, over 1448208.23 frames. ], batch size: 18, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:11:10,233 INFO [optim.py:369] (0/4) 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 11:11:17,242 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.7388, 2.2369, 3.5730, 1.9071, 1.7905, 3.4381, 0.6121, 2.1306], + device='cuda:0'), covar=tensor([0.1375, 0.1013, 0.0205, 0.1516, 0.2096, 0.0320, 0.2099, 0.1409], + device='cuda:0'), in_proj_covar=tensor([0.0196, 0.0203, 0.0133, 0.0221, 0.0274, 0.0144, 0.0172, 0.0197], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 11:11:37,197 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=210650.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:11:39,680 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.6688, 2.9504, 2.2904, 4.0937, 1.6514, 2.2418, 2.7601, 2.9510], + device='cuda:0'), covar=tensor([0.0656, 0.0739, 0.0871, 0.0234, 0.1121, 0.1186, 0.0828, 0.0728], + device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0194, 0.0246, 0.0211, 0.0203, 0.0244, 0.0249, 0.0205], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 11:11:40,169 INFO [train.py:901] (0/4) Epoch 27, batch 500, loss[loss=0.1788, simple_loss=0.2582, pruned_loss=0.04966, over 7822.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2835, pruned_loss=0.05901, over 1484979.82 frames. ], batch size: 20, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:11:54,645 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.01 vs. limit=5.0 +2023-02-07 11:12:01,203 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=210684.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:12:08,765 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4154, 1.4787, 1.3075, 1.6590, 1.0206, 1.2105, 1.4472, 1.5025], + device='cuda:0'), covar=tensor([0.0634, 0.0625, 0.0720, 0.0518, 0.0910, 0.0972, 0.0561, 0.0564], + device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0194, 0.0246, 0.0210, 0.0203, 0.0244, 0.0249, 0.0205], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 11:12:12,805 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=210700.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:12:15,960 INFO [train.py:901] (0/4) Epoch 27, batch 550, loss[loss=0.2217, simple_loss=0.3017, pruned_loss=0.07088, over 8032.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2845, pruned_loss=0.05919, over 1513425.13 frames. ], batch size: 22, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:12:19,369 INFO [optim.py:369] (0/4) 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 11:12:35,240 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5957, 1.4361, 1.6204, 1.3177, 0.7780, 1.3636, 1.3327, 1.3319], + device='cuda:0'), covar=tensor([0.0605, 0.1246, 0.1684, 0.1519, 0.0645, 0.1527, 0.0783, 0.0695], + device='cuda:0'), in_proj_covar=tensor([0.0099, 0.0154, 0.0190, 0.0162, 0.0101, 0.0164, 0.0113, 0.0146], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:0') +2023-02-07 11:12:50,315 INFO [train.py:901] (0/4) Epoch 27, batch 600, loss[loss=0.1897, simple_loss=0.2846, pruned_loss=0.04742, over 8459.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2847, pruned_loss=0.05869, over 1537167.82 frames. ], batch size: 27, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:12:50,560 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6072, 2.1008, 3.2865, 1.5128, 2.5332, 2.0126, 1.8168, 2.6575], + device='cuda:0'), covar=tensor([0.1874, 0.2637, 0.0771, 0.4479, 0.1709, 0.3230, 0.2340, 0.1943], + device='cuda:0'), in_proj_covar=tensor([0.0537, 0.0631, 0.0562, 0.0665, 0.0660, 0.0609, 0.0561, 0.0644], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 11:13:00,512 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.7676, 2.0948, 3.6176, 2.0426, 1.6839, 3.5157, 0.5309, 2.1566], + device='cuda:0'), covar=tensor([0.1505, 0.1603, 0.0338, 0.1526, 0.2606, 0.0376, 0.2381, 0.1372], + device='cuda:0'), in_proj_covar=tensor([0.0197, 0.0203, 0.0133, 0.0221, 0.0276, 0.0144, 0.0172, 0.0198], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 11:13:05,184 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8692, 1.5275, 1.7481, 1.4258, 1.0889, 1.5333, 1.7885, 1.4700], + device='cuda:0'), covar=tensor([0.0550, 0.1256, 0.1638, 0.1415, 0.0595, 0.1446, 0.0685, 0.0681], + device='cuda:0'), in_proj_covar=tensor([0.0099, 0.0154, 0.0190, 0.0162, 0.0101, 0.0164, 0.0113, 0.0146], + device='cuda:0'), out_proj_covar=tensor([0.0007, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:0') +2023-02-07 11:13:07,221 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6111, 1.4706, 1.9768, 1.3602, 1.2608, 1.9477, 0.5995, 1.3233], + device='cuda:0'), covar=tensor([0.1297, 0.1212, 0.0384, 0.0845, 0.2145, 0.0417, 0.1766, 0.1287], + device='cuda:0'), in_proj_covar=tensor([0.0196, 0.0203, 0.0133, 0.0221, 0.0275, 0.0144, 0.0172, 0.0197], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 11:13:08,464 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=210782.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:13:11,593 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465_sp0.9 from training. Duration: 29.816625 +2023-02-07 11:13:13,654 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=210790.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:13:23,577 INFO [train.py:901] (0/4) Epoch 27, batch 650, loss[loss=0.1602, simple_loss=0.2475, pruned_loss=0.03648, over 7713.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2845, pruned_loss=0.0584, over 1558638.50 frames. ], batch size: 18, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:13:28,267 INFO [optim.py:369] (0/4) 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 11:13:32,497 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210815.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:13:59,787 INFO [train.py:901] (0/4) Epoch 27, batch 700, loss[loss=0.2135, simple_loss=0.3031, pruned_loss=0.06198, over 8107.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.284, pruned_loss=0.05829, over 1574135.89 frames. ], batch size: 23, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:14:03,232 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1853, 2.0610, 2.6643, 2.2014, 2.5957, 2.2739, 2.1060, 1.4276], + device='cuda:0'), covar=tensor([0.5690, 0.5024, 0.2075, 0.3745, 0.2496, 0.3132, 0.1947, 0.5341], + device='cuda:0'), in_proj_covar=tensor([0.0961, 0.1015, 0.0827, 0.0983, 0.1018, 0.0925, 0.0768, 0.0845], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 11:14:16,332 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.5522, 1.8118, 5.8368, 2.5688, 4.9253, 4.8848, 5.3683, 5.3287], + device='cuda:0'), covar=tensor([0.0857, 0.6390, 0.0607, 0.4132, 0.1589, 0.1233, 0.0832, 0.0693], + device='cuda:0'), in_proj_covar=tensor([0.0665, 0.0659, 0.0728, 0.0650, 0.0737, 0.0629, 0.0629, 0.0702], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 11:14:32,709 INFO [train.py:901] (0/4) Epoch 27, batch 750, loss[loss=0.2347, simple_loss=0.2962, pruned_loss=0.08656, over 7634.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2857, pruned_loss=0.05954, over 1586806.14 frames. ], batch size: 19, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:14:33,743 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.91 vs. limit=5.0 +2023-02-07 11:14:35,963 INFO [optim.py:369] (0/4) 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 11:14:54,972 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994_sp0.9 from training. Duration: 30.1555625 +2023-02-07 11:15:04,232 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 27, batch 800, loss[loss=0.2036, simple_loss=0.2996, pruned_loss=0.05374, over 8254.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2839, pruned_loss=0.05875, over 1589824.02 frames. ], batch size: 24, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:15:11,720 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.28 vs. limit=5.0 +2023-02-07 11:15:34,982 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=210994.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:15:38,620 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 +2023-02-07 11:15:40,383 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 27, batch 850, loss[loss=0.202, simple_loss=0.288, pruned_loss=0.05799, over 7938.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2835, pruned_loss=0.05857, over 1591712.17 frames. ], batch size: 20, lr: 2.81e-03, grad_scale: 16.0 +2023-02-07 11:15:45,629 INFO [optim.py:369] (0/4) 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 11:15:46,885 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 +2023-02-07 11:15:57,712 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=211028.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:16:10,425 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=211044.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:16:17,675 INFO [train.py:901] (0/4) Epoch 27, batch 900, loss[loss=0.193, simple_loss=0.2853, pruned_loss=0.0504, over 8647.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.284, pruned_loss=0.05922, over 1596229.84 frames. ], batch size: 34, lr: 2.81e-03, grad_scale: 16.0 +2023-02-07 11:16:51,933 INFO [train.py:901] (0/4) Epoch 27, batch 950, loss[loss=0.1828, simple_loss=0.2655, pruned_loss=0.05003, over 7817.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2826, pruned_loss=0.05812, over 1601879.07 frames. ], batch size: 20, lr: 2.81e-03, grad_scale: 16.0 +2023-02-07 11:16:54,833 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 11:17:06,152 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=211126.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:17:11,859 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 +2023-02-07 11:17:17,576 INFO [zipformer.py:1185] (0/4) 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] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp1.1 from training. Duration: 25.3818125 +2023-02-07 11:17:26,814 INFO [train.py:901] (0/4) Epoch 27, batch 1000, loss[loss=0.1983, simple_loss=0.2889, pruned_loss=0.05391, over 8290.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2825, pruned_loss=0.05808, over 1607068.79 frames. ], batch size: 23, lr: 2.81e-03, grad_scale: 16.0 +2023-02-07 11:17:30,456 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=211159.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:17:51,441 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3025, 2.1506, 1.7497, 2.0400, 1.7721, 1.4832, 1.7435, 1.8073], + device='cuda:0'), covar=tensor([0.1221, 0.0409, 0.1199, 0.0496, 0.0769, 0.1555, 0.0925, 0.0682], + device='cuda:0'), in_proj_covar=tensor([0.0360, 0.0239, 0.0340, 0.0311, 0.0304, 0.0344, 0.0346, 0.0321], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 11:17:53,293 WARNING [train.py:1067] (0/4) Exclude cut with ID 6951-79737-0043-83149 from training. Duration: 25.285 +2023-02-07 11:18:03,291 INFO [train.py:901] (0/4) Epoch 27, batch 1050, loss[loss=0.1995, simple_loss=0.2916, pruned_loss=0.05366, over 8573.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2823, pruned_loss=0.05775, over 1610088.96 frames. ], batch size: 39, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:18:05,217 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403 from training. Duration: 29.735 +2023-02-07 11:18:07,251 INFO [optim.py:369] (0/4) 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] (0/4) attn_weights_entropy = tensor([1.4442, 1.2856, 1.5051, 1.3628, 1.4329, 1.4520, 1.3873, 0.7649], + device='cuda:0'), covar=tensor([0.4277, 0.3570, 0.1698, 0.2725, 0.2012, 0.2482, 0.1469, 0.3876], + device='cuda:0'), in_proj_covar=tensor([0.0964, 0.1018, 0.0831, 0.0988, 0.1021, 0.0928, 0.0771, 0.0850], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 11:18:27,161 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=211241.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:18:36,465 INFO [train.py:901] (0/4) Epoch 27, batch 1100, loss[loss=0.2009, simple_loss=0.2943, pruned_loss=0.05369, over 8358.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2828, pruned_loss=0.05814, over 1611491.83 frames. ], batch size: 24, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:19:13,344 INFO [train.py:901] (0/4) Epoch 27, batch 1150, loss[loss=0.2013, simple_loss=0.2673, pruned_loss=0.06768, over 7802.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2825, pruned_loss=0.0579, over 1613027.32 frames. ], batch size: 19, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:19:15,957 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467_sp0.9 from training. Duration: 27.8166875 +2023-02-07 11:19:17,153 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 +2023-02-07 11:19:17,281 INFO [optim.py:369] (0/4) 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 11:19:40,822 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=211346.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:19:46,769 INFO [train.py:901] (0/4) Epoch 27, batch 1200, loss[loss=0.211, simple_loss=0.3029, pruned_loss=0.05958, over 8042.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2819, pruned_loss=0.05764, over 1609024.81 frames. ], batch size: 22, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:19:53,739 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=211365.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:20:08,479 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0469, 2.2124, 1.8468, 2.8329, 1.3301, 1.6403, 2.0654, 2.2030], + device='cuda:0'), covar=tensor([0.0787, 0.0800, 0.0831, 0.0344, 0.1073, 0.1292, 0.0808, 0.0825], + device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0194, 0.0244, 0.0211, 0.0202, 0.0244, 0.0249, 0.0204], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 11:20:11,155 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=211390.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:20:18,555 INFO [zipformer.py:1185] (0/4) 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] (0/4) attn_weights_entropy = tensor([4.7079, 4.7711, 4.2764, 2.0259, 4.1490, 4.3680, 4.3166, 4.1706], + device='cuda:0'), covar=tensor([0.0663, 0.0498, 0.0961, 0.4600, 0.0870, 0.0864, 0.1108, 0.0801], + device='cuda:0'), in_proj_covar=tensor([0.0543, 0.0459, 0.0446, 0.0557, 0.0440, 0.0463, 0.0437, 0.0407], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 11:20:21,762 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=211404.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:20:22,039 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 +2023-02-07 11:20:22,319 INFO [train.py:901] (0/4) Epoch 27, batch 1250, loss[loss=0.2043, simple_loss=0.2897, pruned_loss=0.05947, over 8518.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.283, pruned_loss=0.05829, over 1612695.15 frames. ], batch size: 26, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:20:26,153 INFO [optim.py:369] (0/4) 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 11:20:29,695 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=211415.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:20:35,507 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=211424.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:20:39,463 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=211430.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:20:46,219 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=211440.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:20:49,956 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 +2023-02-07 11:20:56,225 INFO [train.py:901] (0/4) Epoch 27, batch 1300, loss[loss=0.2039, simple_loss=0.2872, pruned_loss=0.06031, over 8469.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2824, pruned_loss=0.05737, over 1617256.02 frames. ], batch size: 25, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:20:59,829 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7082, 2.1187, 3.3823, 1.5730, 2.5196, 2.1868, 1.8803, 2.5854], + device='cuda:0'), covar=tensor([0.1979, 0.2941, 0.0875, 0.4942, 0.1980, 0.3486, 0.2497, 0.2361], + device='cuda:0'), in_proj_covar=tensor([0.0539, 0.0632, 0.0564, 0.0669, 0.0659, 0.0610, 0.0560, 0.0644], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 11:21:00,455 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=211461.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:21:23,158 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0885, 1.6594, 4.1638, 1.8155, 2.4381, 4.7087, 4.8127, 4.0897], + device='cuda:0'), covar=tensor([0.1326, 0.1947, 0.0322, 0.2126, 0.1251, 0.0193, 0.0452, 0.0569], + device='cuda:0'), in_proj_covar=tensor([0.0307, 0.0327, 0.0293, 0.0321, 0.0320, 0.0278, 0.0437, 0.0307], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 11:21:24,561 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=211497.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:21:30,505 INFO [train.py:901] (0/4) Epoch 27, batch 1350, loss[loss=0.2023, simple_loss=0.2906, pruned_loss=0.05706, over 8089.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2812, pruned_loss=0.05671, over 1617169.34 frames. ], batch size: 21, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:21:34,476 INFO [optim.py:369] (0/4) 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 11:21:43,592 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=211522.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:22:05,937 INFO [train.py:901] (0/4) Epoch 27, batch 1400, loss[loss=0.2118, simple_loss=0.2968, pruned_loss=0.0634, over 8502.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2817, pruned_loss=0.05717, over 1620206.28 frames. ], batch size: 26, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:22:39,548 INFO [train.py:901] (0/4) Epoch 27, batch 1450, loss[loss=0.2612, simple_loss=0.3579, pruned_loss=0.0822, over 8354.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2838, pruned_loss=0.05842, over 1620062.84 frames. ], batch size: 24, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:22:43,642 INFO [optim.py:369] (0/4) 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 11:22:45,699 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0007-9590 from training. Duration: 25.85 +2023-02-07 11:23:15,887 INFO [train.py:901] (0/4) Epoch 27, batch 1500, loss[loss=0.1715, simple_loss=0.2635, pruned_loss=0.03971, over 8022.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2839, pruned_loss=0.05847, over 1617780.53 frames. ], batch size: 22, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:23:49,651 INFO [train.py:901] (0/4) Epoch 27, batch 1550, loss[loss=0.2283, simple_loss=0.3053, pruned_loss=0.07562, over 8083.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2836, pruned_loss=0.05834, over 1620335.82 frames. ], batch size: 21, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:23:53,681 INFO [optim.py:369] (0/4) 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] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=211717.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:24:15,282 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=211742.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:24:19,419 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=211748.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:24:24,782 INFO [train.py:901] (0/4) Epoch 27, batch 1600, loss[loss=0.2429, simple_loss=0.3273, pruned_loss=0.0793, over 8494.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2836, pruned_loss=0.0581, over 1624892.35 frames. ], batch size: 28, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:24:28,356 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.0045, 2.3300, 3.7444, 2.1910, 2.0838, 3.7752, 0.7467, 2.2372], + device='cuda:0'), covar=tensor([0.1114, 0.1071, 0.0224, 0.1437, 0.2029, 0.0190, 0.1958, 0.1623], + device='cuda:0'), in_proj_covar=tensor([0.0197, 0.0203, 0.0133, 0.0221, 0.0275, 0.0144, 0.0172, 0.0197], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 11:24:38,870 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=211774.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:24:59,712 INFO [train.py:901] (0/4) Epoch 27, batch 1650, loss[loss=0.1773, simple_loss=0.2575, pruned_loss=0.04856, over 7930.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2835, pruned_loss=0.05767, over 1627846.97 frames. ], batch size: 20, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:25:03,543 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 +2023-02-07 11:25:03,792 INFO [optim.py:369] (0/4) 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 11:25:34,306 INFO [train.py:901] (0/4) Epoch 27, batch 1700, loss[loss=0.2184, simple_loss=0.2902, pruned_loss=0.07332, over 6960.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2843, pruned_loss=0.05787, over 1627214.07 frames. ], batch size: 72, lr: 2.81e-03, grad_scale: 8.0 +2023-02-07 11:25:39,903 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=211863.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:25:57,117 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5885, 1.3888, 2.8784, 1.4351, 2.1517, 3.0994, 3.2436, 2.6359], + device='cuda:0'), covar=tensor([0.1251, 0.1654, 0.0347, 0.2053, 0.0891, 0.0299, 0.0487, 0.0550], + device='cuda:0'), in_proj_covar=tensor([0.0307, 0.0329, 0.0295, 0.0322, 0.0323, 0.0279, 0.0440, 0.0308], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 11:25:59,114 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 27, batch 1750, loss[loss=0.1837, simple_loss=0.2642, pruned_loss=0.05165, over 8241.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2838, pruned_loss=0.05771, over 1624455.22 frames. ], batch size: 22, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:26:13,486 INFO [optim.py:369] (0/4) 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 11:26:24,668 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=4.14 vs. limit=5.0 +2023-02-07 11:26:43,427 INFO [train.py:901] (0/4) Epoch 27, batch 1800, loss[loss=0.2024, simple_loss=0.2904, pruned_loss=0.05726, over 8312.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2832, pruned_loss=0.05731, over 1620426.82 frames. ], batch size: 25, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:27:03,807 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 +2023-02-07 11:27:15,627 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-212000.pt +2023-02-07 11:27:20,494 INFO [train.py:901] (0/4) Epoch 27, batch 1850, loss[loss=0.1948, simple_loss=0.2771, pruned_loss=0.05628, over 8294.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2816, pruned_loss=0.0568, over 1615436.96 frames. ], batch size: 23, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:27:24,576 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 27, batch 1900, loss[loss=0.1787, simple_loss=0.2632, pruned_loss=0.04704, over 7417.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2832, pruned_loss=0.05784, over 1616803.89 frames. ], batch size: 17, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:28:23,227 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.6275, 4.5968, 4.2249, 2.0898, 4.0829, 4.2141, 4.1798, 4.0971], + device='cuda:0'), covar=tensor([0.0611, 0.0484, 0.0910, 0.4038, 0.0771, 0.0862, 0.1071, 0.0667], + device='cuda:0'), in_proj_covar=tensor([0.0540, 0.0456, 0.0442, 0.0552, 0.0438, 0.0460, 0.0437, 0.0402], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 11:28:25,173 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0021-44397_sp0.9 from training. Duration: 27.511125 +2023-02-07 11:28:28,373 INFO [train.py:901] (0/4) Epoch 27, batch 1950, loss[loss=0.1779, simple_loss=0.2677, pruned_loss=0.04403, over 8251.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2829, pruned_loss=0.05752, over 1615400.39 frames. ], batch size: 24, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:28:33,122 INFO [optim.py:369] (0/4) 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 11:28:38,408 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390 from training. Duration: 27.92 +2023-02-07 11:28:38,624 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=212119.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:28:41,246 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=212122.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 11:28:56,769 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=212144.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:28:57,301 WARNING [train.py:1067] (0/4) Exclude cut with ID 4964-30587-0040-138716_sp0.9 from training. Duration: 25.0944375 +2023-02-07 11:28:57,488 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=212145.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:29:03,980 INFO [train.py:901] (0/4) Epoch 27, batch 2000, loss[loss=0.1953, simple_loss=0.2808, pruned_loss=0.05495, over 8525.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2823, pruned_loss=0.057, over 1614377.57 frames. ], batch size: 28, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:29:14,165 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=212170.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:29:37,643 INFO [train.py:901] (0/4) Epoch 27, batch 2050, loss[loss=0.2352, simple_loss=0.3211, pruned_loss=0.07467, over 7100.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2829, pruned_loss=0.0575, over 1613558.74 frames. ], batch size: 72, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:29:41,746 INFO [optim.py:369] (0/4) 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 11:29:53,464 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6304, 1.4347, 1.6398, 1.3635, 0.9363, 1.4415, 1.4495, 1.3491], + device='cuda:0'), covar=tensor([0.0613, 0.1284, 0.1759, 0.1522, 0.0616, 0.1540, 0.0745, 0.0706], + device='cuda:0'), in_proj_covar=tensor([0.0099, 0.0154, 0.0190, 0.0163, 0.0102, 0.0164, 0.0113, 0.0146], + device='cuda:0'), out_proj_covar=tensor([0.0007, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:0') +2023-02-07 11:30:13,814 INFO [train.py:901] (0/4) Epoch 27, batch 2100, loss[loss=0.2111, simple_loss=0.2979, pruned_loss=0.06215, over 8111.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2816, pruned_loss=0.05693, over 1607408.67 frames. ], batch size: 23, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:30:24,393 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.4056, 4.3448, 4.0014, 1.9425, 3.8390, 4.0657, 3.8951, 3.9037], + device='cuda:0'), covar=tensor([0.0693, 0.0501, 0.0946, 0.4576, 0.0860, 0.0978, 0.1240, 0.0732], + device='cuda:0'), in_proj_covar=tensor([0.0539, 0.0456, 0.0442, 0.0554, 0.0438, 0.0459, 0.0437, 0.0404], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 11:30:25,136 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=212271.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:30:42,128 INFO [scaling.py:679] (0/4) 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] (0/4) Epoch 27, batch 2150, loss[loss=0.2275, simple_loss=0.306, pruned_loss=0.07453, over 8502.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.281, pruned_loss=0.05639, over 1610343.72 frames. ], batch size: 26, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:30:48,019 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([0.9711, 1.5974, 1.3967, 1.4901, 1.2659, 1.2571, 1.2147, 1.2527], + device='cuda:0'), covar=tensor([0.1303, 0.0602, 0.1429, 0.0635, 0.0835, 0.1710, 0.1056, 0.0888], + device='cuda:0'), in_proj_covar=tensor([0.0365, 0.0240, 0.0341, 0.0316, 0.0304, 0.0349, 0.0350, 0.0323], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 11:30:51,761 INFO [optim.py:369] (0/4) 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 11:31:22,777 INFO [train.py:901] (0/4) Epoch 27, batch 2200, loss[loss=0.1779, simple_loss=0.2593, pruned_loss=0.04827, over 7798.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2813, pruned_loss=0.05683, over 1611416.88 frames. ], batch size: 20, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:31:57,346 INFO [train.py:901] (0/4) Epoch 27, batch 2250, loss[loss=0.1879, simple_loss=0.273, pruned_loss=0.05141, over 7734.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2822, pruned_loss=0.05671, over 1613143.19 frames. ], batch size: 18, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:32:01,577 INFO [optim.py:369] (0/4) 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 11:32:31,466 INFO [train.py:901] (0/4) Epoch 27, batch 2300, loss[loss=0.1583, simple_loss=0.2506, pruned_loss=0.03299, over 7813.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2814, pruned_loss=0.05611, over 1614194.61 frames. ], batch size: 20, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:32:32,832 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=212457.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:32:39,375 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=212466.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 11:33:07,947 INFO [train.py:901] (0/4) Epoch 27, batch 2350, loss[loss=0.1978, simple_loss=0.2815, pruned_loss=0.05698, over 8033.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.2807, pruned_loss=0.05585, over 1612593.85 frames. ], batch size: 22, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:33:12,149 INFO [optim.py:369] (0/4) 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 11:33:42,857 INFO [train.py:901] (0/4) Epoch 27, batch 2400, loss[loss=0.1862, simple_loss=0.262, pruned_loss=0.05521, over 7657.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2816, pruned_loss=0.05687, over 1609670.24 frames. ], batch size: 19, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:34:01,662 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=212581.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 11:34:04,897 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.8377, 3.7892, 3.4452, 1.9415, 3.3699, 3.5654, 3.3453, 3.3523], + device='cuda:0'), covar=tensor([0.0940, 0.0712, 0.1215, 0.4548, 0.0989, 0.1127, 0.1519, 0.0887], + device='cuda:0'), in_proj_covar=tensor([0.0542, 0.0460, 0.0444, 0.0558, 0.0440, 0.0464, 0.0439, 0.0406], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 11:34:19,817 INFO [train.py:901] (0/4) Epoch 27, batch 2450, loss[loss=0.2055, simple_loss=0.2995, pruned_loss=0.05581, over 8504.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.282, pruned_loss=0.05672, over 1615834.65 frames. ], batch size: 28, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:34:23,907 INFO [optim.py:369] (0/4) 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 11:34:26,770 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=212615.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:34:46,275 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 +2023-02-07 11:34:54,050 INFO [train.py:901] (0/4) Epoch 27, batch 2500, loss[loss=0.2016, simple_loss=0.2819, pruned_loss=0.06063, over 7231.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2831, pruned_loss=0.0577, over 1617890.82 frames. ], batch size: 16, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:34:59,673 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8285, 1.6440, 3.3703, 1.4132, 2.3487, 3.6248, 3.7347, 3.1387], + device='cuda:0'), covar=tensor([0.1326, 0.1787, 0.0334, 0.2347, 0.1047, 0.0226, 0.0564, 0.0520], + device='cuda:0'), in_proj_covar=tensor([0.0305, 0.0327, 0.0292, 0.0321, 0.0320, 0.0277, 0.0437, 0.0308], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 11:35:28,284 INFO [train.py:901] (0/4) Epoch 27, batch 2550, loss[loss=0.1542, simple_loss=0.2352, pruned_loss=0.0366, over 7699.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2835, pruned_loss=0.05809, over 1618970.20 frames. ], batch size: 18, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:35:33,073 INFO [optim.py:369] (0/4) 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 11:35:37,328 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=212716.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:35:46,863 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=212730.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:36:04,558 INFO [train.py:901] (0/4) Epoch 27, batch 2600, loss[loss=0.2037, simple_loss=0.2885, pruned_loss=0.05941, over 8565.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2836, pruned_loss=0.05783, over 1616865.05 frames. ], batch size: 31, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:36:09,617 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 +2023-02-07 11:36:18,100 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9663, 1.6604, 3.5101, 1.6431, 2.6047, 3.8183, 3.8758, 3.3285], + device='cuda:0'), covar=tensor([0.1190, 0.1722, 0.0290, 0.2028, 0.0898, 0.0199, 0.0501, 0.0468], + device='cuda:0'), in_proj_covar=tensor([0.0305, 0.0327, 0.0292, 0.0320, 0.0320, 0.0277, 0.0438, 0.0307], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 11:36:21,569 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1354, 3.5572, 2.2000, 2.7082, 2.8463, 1.9341, 2.8587, 2.9594], + device='cuda:0'), covar=tensor([0.1864, 0.0475, 0.1306, 0.0883, 0.0800, 0.1590, 0.1163, 0.1305], + device='cuda:0'), in_proj_covar=tensor([0.0359, 0.0238, 0.0337, 0.0312, 0.0301, 0.0345, 0.0346, 0.0319], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 11:36:29,683 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=212792.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:36:33,818 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.0236, 2.2751, 3.7934, 2.2446, 2.0953, 3.7502, 0.8078, 2.3097], + device='cuda:0'), covar=tensor([0.1185, 0.1315, 0.0182, 0.1243, 0.2086, 0.0234, 0.1999, 0.1360], + device='cuda:0'), in_proj_covar=tensor([0.0198, 0.0205, 0.0134, 0.0223, 0.0276, 0.0144, 0.0171, 0.0198], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 11:36:35,822 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=212801.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:36:38,497 INFO [train.py:901] (0/4) Epoch 27, batch 2650, loss[loss=0.2089, simple_loss=0.2906, pruned_loss=0.0636, over 8449.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2826, pruned_loss=0.05766, over 1614781.39 frames. ], batch size: 27, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:36:43,316 INFO [optim.py:369] (0/4) 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 11:37:02,114 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=212837.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 11:37:04,071 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5796, 1.4824, 1.8860, 1.2763, 1.2070, 1.8576, 0.1805, 1.2514], + device='cuda:0'), covar=tensor([0.1305, 0.1180, 0.0365, 0.0843, 0.2165, 0.0421, 0.1728, 0.1197], + device='cuda:0'), in_proj_covar=tensor([0.0198, 0.0205, 0.0134, 0.0223, 0.0276, 0.0145, 0.0172, 0.0198], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 11:37:14,955 INFO [train.py:901] (0/4) Epoch 27, batch 2700, loss[loss=0.2262, simple_loss=0.3073, pruned_loss=0.07249, over 8322.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2839, pruned_loss=0.0588, over 1612927.48 frames. ], batch size: 25, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:37:19,841 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=212862.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 11:37:49,368 INFO [train.py:901] (0/4) Epoch 27, batch 2750, loss[loss=0.2098, simple_loss=0.297, pruned_loss=0.06129, over 8484.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2833, pruned_loss=0.05896, over 1609976.64 frames. ], batch size: 39, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:37:53,340 INFO [optim.py:369] (0/4) 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 11:37:56,793 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=212916.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:38:25,581 INFO [train.py:901] (0/4) Epoch 27, batch 2800, loss[loss=0.2049, simple_loss=0.2911, pruned_loss=0.05933, over 8279.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2846, pruned_loss=0.05946, over 1612441.27 frames. ], batch size: 23, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:38:43,385 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4880, 1.4772, 1.8395, 1.1272, 1.0575, 1.8502, 0.1932, 1.1840], + device='cuda:0'), covar=tensor([0.1605, 0.1022, 0.0393, 0.1080, 0.2603, 0.0387, 0.1676, 0.1216], + device='cuda:0'), in_proj_covar=tensor([0.0199, 0.0205, 0.0134, 0.0224, 0.0277, 0.0145, 0.0172, 0.0199], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 11:38:46,070 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=212986.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:38:58,528 INFO [train.py:901] (0/4) Epoch 27, batch 2850, loss[loss=0.2533, simple_loss=0.3451, pruned_loss=0.08075, over 8513.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2851, pruned_loss=0.05966, over 1612000.87 frames. ], batch size: 49, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:39:02,628 INFO [optim.py:369] (0/4) 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 11:39:02,841 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=213011.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:39:03,500 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5608, 1.4904, 1.7637, 1.3412, 0.8720, 1.5115, 1.4740, 1.1784], + device='cuda:0'), covar=tensor([0.0608, 0.1247, 0.1565, 0.1503, 0.0603, 0.1451, 0.0718, 0.0749], + device='cuda:0'), in_proj_covar=tensor([0.0099, 0.0153, 0.0189, 0.0162, 0.0101, 0.0164, 0.0112, 0.0146], + device='cuda:0'), out_proj_covar=tensor([0.0007, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:0') +2023-02-07 11:39:33,429 INFO [train.py:901] (0/4) Epoch 27, batch 2900, loss[loss=0.2269, simple_loss=0.3107, pruned_loss=0.07161, over 8332.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2846, pruned_loss=0.05928, over 1610345.37 frames. ], batch size: 25, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:39:36,850 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=213060.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:39:46,822 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=213073.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:40:08,834 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.89 vs. limit=5.0 +2023-02-07 11:40:09,162 INFO [train.py:901] (0/4) Epoch 27, batch 2950, loss[loss=0.1913, simple_loss=0.288, pruned_loss=0.04732, over 8448.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2836, pruned_loss=0.05869, over 1608741.75 frames. ], batch size: 29, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:40:12,522 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp1.1 from training. Duration: 0.7545625 +2023-02-07 11:40:13,188 INFO [optim.py:369] (0/4) 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 11:40:30,262 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 27, batch 3000, loss[loss=0.2177, simple_loss=0.307, pruned_loss=0.06422, over 8197.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.284, pruned_loss=0.05852, over 1610470.36 frames. ], batch size: 23, lr: 2.80e-03, grad_scale: 8.0 +2023-02-07 11:40:42,823 INFO [train.py:926] (0/4) Computing validation loss +2023-02-07 11:40:56,481 INFO [train.py:935] (0/4) Epoch 27, validation: loss=0.171, simple_loss=0.2706, pruned_loss=0.03572, over 944034.00 frames. +2023-02-07 11:40:56,482 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6760MB +2023-02-07 11:41:08,332 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=213172.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:41:10,342 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=213175.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:41:19,861 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=213188.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:41:25,937 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 27, batch 3050, loss[loss=0.2481, simple_loss=0.3175, pruned_loss=0.08931, over 7238.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.283, pruned_loss=0.0582, over 1608695.47 frames. ], batch size: 71, lr: 2.80e-03, grad_scale: 16.0 +2023-02-07 11:41:36,535 INFO [optim.py:369] (0/4) 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 11:41:49,446 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4775, 1.7435, 4.3420, 1.9281, 2.6032, 4.9800, 5.1256, 4.2769], + device='cuda:0'), covar=tensor([0.1104, 0.1848, 0.0272, 0.2029, 0.1190, 0.0183, 0.0431, 0.0559], + device='cuda:0'), in_proj_covar=tensor([0.0306, 0.0328, 0.0293, 0.0321, 0.0321, 0.0279, 0.0438, 0.0309], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 11:41:55,419 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.2358, 1.5299, 4.5201, 1.9190, 3.8251, 3.6650, 4.1669, 4.0997], + device='cuda:0'), covar=tensor([0.1010, 0.6102, 0.0952, 0.4674, 0.1732, 0.1511, 0.0871, 0.0826], + device='cuda:0'), in_proj_covar=tensor([0.0676, 0.0668, 0.0736, 0.0660, 0.0749, 0.0637, 0.0638, 0.0719], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 11:42:04,098 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=213251.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:42:06,427 INFO [train.py:901] (0/4) Epoch 27, batch 3100, loss[loss=0.2165, simple_loss=0.2963, pruned_loss=0.06831, over 8032.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2834, pruned_loss=0.0586, over 1607073.05 frames. ], batch size: 22, lr: 2.80e-03, grad_scale: 16.0 +2023-02-07 11:42:40,166 INFO [train.py:901] (0/4) Epoch 27, batch 3150, loss[loss=0.182, simple_loss=0.2655, pruned_loss=0.04927, over 8089.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2834, pruned_loss=0.0588, over 1608142.98 frames. ], batch size: 21, lr: 2.80e-03, grad_scale: 16.0 +2023-02-07 11:42:44,217 INFO [optim.py:369] (0/4) 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] (0/4) Epoch 27, batch 3200, loss[loss=0.1895, simple_loss=0.2857, pruned_loss=0.04665, over 8505.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2836, pruned_loss=0.05841, over 1613844.27 frames. ], batch size: 28, lr: 2.80e-03, grad_scale: 16.0 +2023-02-07 11:43:48,888 INFO [train.py:901] (0/4) Epoch 27, batch 3250, loss[loss=0.228, simple_loss=0.3089, pruned_loss=0.07359, over 8289.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2828, pruned_loss=0.05766, over 1611684.38 frames. ], batch size: 49, lr: 2.80e-03, grad_scale: 16.0 +2023-02-07 11:43:52,809 INFO [optim.py:369] (0/4) 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 11:44:07,271 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=213431.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:44:17,673 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=213444.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:44:25,453 INFO [train.py:901] (0/4) Epoch 27, batch 3300, loss[loss=0.2204, simple_loss=0.2842, pruned_loss=0.07834, over 7793.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2825, pruned_loss=0.05763, over 1612080.98 frames. ], batch size: 19, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:44:26,254 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=213456.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:44:35,188 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=213469.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:44:59,365 INFO [train.py:901] (0/4) Epoch 27, batch 3350, loss[loss=0.2508, simple_loss=0.3352, pruned_loss=0.08314, over 8718.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2825, pruned_loss=0.05802, over 1610632.53 frames. ], batch size: 30, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:45:00,955 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 11:45:18,435 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=213532.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:45:33,327 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4250, 1.4287, 1.3968, 1.8418, 0.5647, 1.2903, 1.2995, 1.4441], + device='cuda:0'), covar=tensor([0.0871, 0.0859, 0.0963, 0.0476, 0.1263, 0.1411, 0.0783, 0.0747], + device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0196, 0.0246, 0.0214, 0.0205, 0.0248, 0.0251, 0.0205], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0006, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 11:45:34,516 INFO [train.py:901] (0/4) Epoch 27, batch 3400, loss[loss=0.1845, simple_loss=0.268, pruned_loss=0.05053, over 8194.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2827, pruned_loss=0.05821, over 1611101.39 frames. ], batch size: 23, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:45:55,328 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=213584.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:46:09,634 INFO [train.py:901] (0/4) Epoch 27, batch 3450, loss[loss=0.1972, simple_loss=0.2864, pruned_loss=0.05397, over 8483.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2827, pruned_loss=0.05778, over 1612753.08 frames. ], batch size: 29, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:46:13,704 INFO [optim.py:369] (0/4) 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 11:46:41,460 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 +2023-02-07 11:46:44,525 INFO [train.py:901] (0/4) Epoch 27, batch 3500, loss[loss=0.2425, simple_loss=0.3228, pruned_loss=0.0811, over 6848.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2833, pruned_loss=0.05804, over 1613094.60 frames. ], batch size: 71, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:47:02,336 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.8940, 2.1498, 3.4451, 2.0356, 2.0169, 3.4590, 0.9413, 2.1312], + device='cuda:0'), covar=tensor([0.1150, 0.1277, 0.0372, 0.1662, 0.2192, 0.0330, 0.1848, 0.1447], + device='cuda:0'), in_proj_covar=tensor([0.0200, 0.0205, 0.0135, 0.0224, 0.0276, 0.0145, 0.0172, 0.0199], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 11:47:08,783 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 +2023-02-07 11:47:11,017 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 27, batch 3550, loss[loss=0.1784, simple_loss=0.2659, pruned_loss=0.04549, over 8133.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2821, pruned_loss=0.05765, over 1612243.43 frames. ], batch size: 22, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:47:24,358 INFO [optim.py:369] (0/4) 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 11:47:55,123 INFO [train.py:901] (0/4) Epoch 27, batch 3600, loss[loss=0.2298, simple_loss=0.3087, pruned_loss=0.07545, over 8503.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2824, pruned_loss=0.05757, over 1614659.81 frames. ], batch size: 26, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:47:58,956 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 +2023-02-07 11:48:31,565 INFO [train.py:901] (0/4) Epoch 27, batch 3650, loss[loss=0.194, simple_loss=0.2691, pruned_loss=0.05947, over 8136.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2823, pruned_loss=0.05748, over 1614250.06 frames. ], batch size: 22, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:48:35,651 INFO [optim.py:369] (0/4) 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 11:49:05,225 INFO [train.py:901] (0/4) Epoch 27, batch 3700, loss[loss=0.1835, simple_loss=0.2581, pruned_loss=0.05446, over 7548.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2827, pruned_loss=0.05792, over 1612627.83 frames. ], batch size: 18, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:49:07,787 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 +2023-02-07 11:49:11,332 WARNING [train.py:1067] (0/4) Exclude cut with ID 2411-132532-0017-25057_sp1.1 from training. Duration: 0.9681875 +2023-02-07 11:49:40,753 INFO [train.py:901] (0/4) Epoch 27, batch 3750, loss[loss=0.213, simple_loss=0.2986, pruned_loss=0.06366, over 8245.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2826, pruned_loss=0.05812, over 1613586.95 frames. ], batch size: 24, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:49:44,686 INFO [optim.py:369] (0/4) 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] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=213928.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:50:03,488 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2941, 3.7146, 2.4851, 3.0521, 3.1181, 2.1893, 3.2046, 3.1761], + device='cuda:0'), covar=tensor([0.1765, 0.0420, 0.1093, 0.0737, 0.0852, 0.1536, 0.0942, 0.1067], + device='cuda:0'), in_proj_covar=tensor([0.0365, 0.0243, 0.0345, 0.0317, 0.0308, 0.0352, 0.0352, 0.0327], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 11:50:15,099 INFO [train.py:901] (0/4) Epoch 27, batch 3800, loss[loss=0.1909, simple_loss=0.271, pruned_loss=0.05539, over 8078.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2833, pruned_loss=0.0584, over 1615721.98 frames. ], batch size: 21, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:50:32,925 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 +2023-02-07 11:50:46,486 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-214000.pt +2023-02-07 11:50:51,694 INFO [train.py:901] (0/4) Epoch 27, batch 3850, loss[loss=0.1961, simple_loss=0.2776, pruned_loss=0.05727, over 8328.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2819, pruned_loss=0.05753, over 1611532.93 frames. ], batch size: 25, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:50:55,388 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 +2023-02-07 11:50:55,678 INFO [optim.py:369] (0/4) 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 11:51:02,708 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0816, 1.2355, 1.2108, 0.7774, 1.2046, 1.0740, 0.1045, 1.2235], + device='cuda:0'), covar=tensor([0.0500, 0.0467, 0.0409, 0.0623, 0.0535, 0.1065, 0.0969, 0.0383], + device='cuda:0'), in_proj_covar=tensor([0.0471, 0.0407, 0.0361, 0.0457, 0.0392, 0.0545, 0.0402, 0.0436], + device='cuda:0'), out_proj_covar=tensor([1.2487e-04, 1.0563e-04, 9.4230e-05, 1.1962e-04, 1.0261e-04, 1.5214e-04, + 1.0725e-04, 1.1451e-04], device='cuda:0') +2023-02-07 11:51:19,438 INFO [zipformer.py:1185] (0/4) 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] (0/4) Exclude cut with ID 3557-8342-0013-71585_sp1.1 from training. Duration: 0.836375 +2023-02-07 11:51:27,202 INFO [train.py:901] (0/4) Epoch 27, batch 3900, loss[loss=0.148, simple_loss=0.229, pruned_loss=0.03346, over 7711.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2828, pruned_loss=0.05774, over 1611216.10 frames. ], batch size: 18, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:51:38,678 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.0574, 1.6817, 3.3407, 1.5738, 2.5144, 3.6448, 3.7610, 3.1720], + device='cuda:0'), covar=tensor([0.1140, 0.1726, 0.0267, 0.2072, 0.0922, 0.0222, 0.0502, 0.0460], + device='cuda:0'), in_proj_covar=tensor([0.0306, 0.0326, 0.0291, 0.0321, 0.0321, 0.0278, 0.0438, 0.0308], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 11:52:00,516 INFO [train.py:901] (0/4) Epoch 27, batch 3950, loss[loss=0.1711, simple_loss=0.2692, pruned_loss=0.03652, over 8183.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2832, pruned_loss=0.05799, over 1614349.97 frames. ], batch size: 23, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:52:04,368 INFO [optim.py:369] (0/4) 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 11:52:26,022 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6290, 1.4361, 1.7633, 1.3113, 0.9253, 1.4980, 1.5119, 1.5059], + device='cuda:0'), covar=tensor([0.0586, 0.1323, 0.1594, 0.1492, 0.0606, 0.1546, 0.0711, 0.0632], + device='cuda:0'), in_proj_covar=tensor([0.0100, 0.0154, 0.0190, 0.0162, 0.0101, 0.0164, 0.0113, 0.0145], + device='cuda:0'), out_proj_covar=tensor([0.0007, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:0') +2023-02-07 11:52:26,775 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9852, 1.7028, 2.0344, 1.8552, 1.9902, 2.0371, 1.9169, 0.8560], + device='cuda:0'), covar=tensor([0.5990, 0.5104, 0.2136, 0.3761, 0.2563, 0.3397, 0.2024, 0.5356], + device='cuda:0'), in_proj_covar=tensor([0.0961, 0.1017, 0.0829, 0.0986, 0.1020, 0.0925, 0.0767, 0.0848], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 11:52:30,026 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=214147.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:52:36,018 INFO [train.py:901] (0/4) Epoch 27, batch 4000, loss[loss=0.1832, simple_loss=0.2616, pruned_loss=0.05245, over 7246.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2829, pruned_loss=0.05794, over 1612644.09 frames. ], batch size: 16, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:52:43,219 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 +2023-02-07 11:53:10,789 INFO [train.py:901] (0/4) Epoch 27, batch 4050, loss[loss=0.1884, simple_loss=0.2637, pruned_loss=0.05656, over 7282.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2831, pruned_loss=0.058, over 1609369.97 frames. ], batch size: 16, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:53:14,931 INFO [optim.py:369] (0/4) 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 11:53:22,008 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-07 11:53:31,859 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=214236.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:53:45,405 INFO [train.py:901] (0/4) Epoch 27, batch 4100, loss[loss=0.1867, simple_loss=0.2759, pruned_loss=0.04874, over 8709.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2814, pruned_loss=0.05724, over 1607376.18 frames. ], batch size: 39, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:54:10,634 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 +2023-02-07 11:54:17,255 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=214299.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:54:21,095 INFO [train.py:901] (0/4) Epoch 27, batch 4150, loss[loss=0.2214, simple_loss=0.3085, pruned_loss=0.06715, over 8444.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2814, pruned_loss=0.05715, over 1608548.62 frames. ], batch size: 24, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:54:25,175 INFO [optim.py:369] (0/4) 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 11:54:34,216 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=214324.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:54:55,411 INFO [train.py:901] (0/4) Epoch 27, batch 4200, loss[loss=0.2493, simple_loss=0.3291, pruned_loss=0.08476, over 8334.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2825, pruned_loss=0.05757, over 1608548.17 frames. ], batch size: 26, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:55:03,017 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.4512, 1.8164, 2.5974, 1.3416, 1.8818, 1.8236, 1.5275, 1.8834], + device='cuda:0'), covar=tensor([0.2062, 0.2705, 0.0983, 0.4905, 0.2170, 0.3426, 0.2648, 0.2496], + device='cuda:0'), in_proj_covar=tensor([0.0540, 0.0635, 0.0566, 0.0671, 0.0660, 0.0609, 0.0564, 0.0647], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 11:55:15,711 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0005-9467 from training. Duration: 25.035 +2023-02-07 11:55:30,763 INFO [train.py:901] (0/4) Epoch 27, batch 4250, loss[loss=0.1824, simple_loss=0.2704, pruned_loss=0.04724, over 7910.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2825, pruned_loss=0.05773, over 1609750.83 frames. ], batch size: 20, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:55:34,789 INFO [optim.py:369] (0/4) 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 11:55:40,121 WARNING [train.py:1067] (0/4) Exclude cut with ID 1914-133440-0024-53073_sp0.9 from training. Duration: 25.2444375 +2023-02-07 11:55:49,286 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.8053, 1.4512, 1.6979, 1.3652, 0.9774, 1.4772, 1.6608, 1.4319], + device='cuda:0'), covar=tensor([0.0580, 0.1332, 0.1673, 0.1500, 0.0587, 0.1551, 0.0717, 0.0704], + device='cuda:0'), in_proj_covar=tensor([0.0099, 0.0154, 0.0190, 0.0161, 0.0101, 0.0163, 0.0112, 0.0145], + device='cuda:0'), out_proj_covar=tensor([0.0007, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:0') +2023-02-07 11:56:04,676 INFO [train.py:901] (0/4) Epoch 27, batch 4300, loss[loss=0.1855, simple_loss=0.2745, pruned_loss=0.04819, over 8137.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2831, pruned_loss=0.05839, over 1609255.38 frames. ], batch size: 22, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:56:10,730 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7063, 2.4083, 4.0017, 1.6080, 2.9789, 2.2704, 1.8833, 3.0188], + device='cuda:0'), covar=tensor([0.2095, 0.2735, 0.0934, 0.4893, 0.1965, 0.3505, 0.2575, 0.2549], + device='cuda:0'), in_proj_covar=tensor([0.0541, 0.0635, 0.0566, 0.0670, 0.0661, 0.0611, 0.0564, 0.0647], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 11:56:23,664 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 +2023-02-07 11:56:29,283 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 27, batch 4350, loss[loss=0.2011, simple_loss=0.2708, pruned_loss=0.06567, over 7800.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2848, pruned_loss=0.05888, over 1618825.74 frames. ], batch size: 19, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:56:44,887 INFO [optim.py:369] (0/4) 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 11:56:48,049 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.89 vs. limit=5.0 +2023-02-07 11:57:11,508 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425_sp0.9 from training. Duration: 28.638875 +2023-02-07 11:57:14,890 INFO [train.py:901] (0/4) Epoch 27, batch 4400, loss[loss=0.2431, simple_loss=0.3147, pruned_loss=0.08574, over 8595.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2839, pruned_loss=0.05853, over 1614485.14 frames. ], batch size: 34, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:57:32,004 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=214580.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:57:39,143 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 +2023-02-07 11:57:49,418 INFO [train.py:901] (0/4) Epoch 27, batch 4450, loss[loss=0.2602, simple_loss=0.3207, pruned_loss=0.09989, over 7914.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2841, pruned_loss=0.05881, over 1614647.88 frames. ], batch size: 20, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:57:50,287 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=214606.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:57:51,442 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp0.9 from training. Duration: 33.038875 +2023-02-07 11:57:53,336 INFO [optim.py:369] (0/4) 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 11:58:06,357 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([5.7840, 1.8854, 5.9899, 2.2530, 5.3669, 5.0426, 5.4656, 5.3822], + device='cuda:0'), covar=tensor([0.0499, 0.4938, 0.0376, 0.3998, 0.1029, 0.0888, 0.0536, 0.0540], + device='cuda:0'), in_proj_covar=tensor([0.0680, 0.0674, 0.0742, 0.0660, 0.0754, 0.0642, 0.0643, 0.0721], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 11:58:10,645 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 +2023-02-07 11:58:14,562 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5788, 2.1879, 3.9844, 1.5429, 2.8989, 2.3808, 1.6456, 2.8318], + device='cuda:0'), covar=tensor([0.2209, 0.2887, 0.0888, 0.5111, 0.2006, 0.3177, 0.2833, 0.2578], + device='cuda:0'), in_proj_covar=tensor([0.0536, 0.0630, 0.0561, 0.0665, 0.0656, 0.0605, 0.0560, 0.0640], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 11:58:25,065 INFO [train.py:901] (0/4) Epoch 27, batch 4500, loss[loss=0.1846, simple_loss=0.2558, pruned_loss=0.05671, over 7537.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2829, pruned_loss=0.05826, over 1613121.13 frames. ], batch size: 18, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:58:49,088 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983 from training. Duration: 0.83 +2023-02-07 11:58:51,961 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=214695.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 11:58:58,558 INFO [train.py:901] (0/4) Epoch 27, batch 4550, loss[loss=0.1619, simple_loss=0.2529, pruned_loss=0.03551, over 8043.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2808, pruned_loss=0.05744, over 1610327.32 frames. ], batch size: 22, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 11:59:03,197 INFO [optim.py:369] (0/4) 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 11:59:34,567 INFO [train.py:901] (0/4) Epoch 27, batch 4600, loss[loss=0.1833, simple_loss=0.2812, pruned_loss=0.04272, over 8251.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2811, pruned_loss=0.05741, over 1610236.59 frames. ], batch size: 24, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 12:00:07,864 INFO [train.py:901] (0/4) Epoch 27, batch 4650, loss[loss=0.1845, simple_loss=0.2551, pruned_loss=0.05693, over 7417.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2819, pruned_loss=0.05825, over 1612888.54 frames. ], batch size: 17, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 12:00:11,921 INFO [optim.py:369] (0/4) 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 12:00:43,657 INFO [train.py:901] (0/4) Epoch 27, batch 4700, loss[loss=0.2378, simple_loss=0.314, pruned_loss=0.08084, over 6789.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2829, pruned_loss=0.0588, over 1609591.63 frames. ], batch size: 71, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 12:00:48,570 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=214862.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:01:05,450 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=214887.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:01:17,189 INFO [train.py:901] (0/4) Epoch 27, batch 4750, loss[loss=0.2396, simple_loss=0.3338, pruned_loss=0.07267, over 8361.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2824, pruned_loss=0.0586, over 1610016.09 frames. ], batch size: 24, lr: 2.79e-03, grad_scale: 16.0 +2023-02-07 12:01:21,120 INFO [optim.py:369] (0/4) 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 12:01:42,271 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0003-9465 from training. Duration: 26.8349375 +2023-02-07 12:01:45,082 WARNING [train.py:1067] (0/4) Exclude cut with ID 5622-44585-0006-50425 from training. Duration: 25.775 +2023-02-07 12:01:49,020 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 27, batch 4800, loss[loss=0.1649, simple_loss=0.2604, pruned_loss=0.03467, over 7809.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2813, pruned_loss=0.05754, over 1608175.44 frames. ], batch size: 20, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:02:07,893 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 27, batch 4850, loss[loss=0.1964, simple_loss=0.2623, pruned_loss=0.06528, over 7532.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2816, pruned_loss=0.05754, over 1608111.61 frames. ], batch size: 18, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:02:31,212 INFO [optim.py:369] (0/4) 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 12:02:36,650 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914 from training. Duration: 26.205 +2023-02-07 12:03:00,128 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9161, 1.9221, 3.2275, 2.3891, 2.7470, 2.0108, 1.6987, 1.6745], + device='cuda:0'), covar=tensor([0.8297, 0.7217, 0.2425, 0.4518, 0.3579, 0.4615, 0.3244, 0.6400], + device='cuda:0'), in_proj_covar=tensor([0.0963, 0.1016, 0.0827, 0.0988, 0.1021, 0.0925, 0.0767, 0.0847], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 12:03:01,830 INFO [train.py:901] (0/4) Epoch 27, batch 4900, loss[loss=0.197, simple_loss=0.2853, pruned_loss=0.05433, over 8015.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.281, pruned_loss=0.05739, over 1605079.87 frames. ], batch size: 22, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:03:37,419 INFO [train.py:901] (0/4) Epoch 27, batch 4950, loss[loss=0.1703, simple_loss=0.2533, pruned_loss=0.04365, over 7415.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.2806, pruned_loss=0.05708, over 1603009.88 frames. ], batch size: 17, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:03:41,324 INFO [optim.py:369] (0/4) 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 12:04:10,562 INFO [train.py:901] (0/4) Epoch 27, batch 5000, loss[loss=0.2013, simple_loss=0.2881, pruned_loss=0.05724, over 8446.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.283, pruned_loss=0.05805, over 1609007.04 frames. ], batch size: 27, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:04:47,121 INFO [train.py:901] (0/4) Epoch 27, batch 5050, loss[loss=0.2768, simple_loss=0.3428, pruned_loss=0.1054, over 7074.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2818, pruned_loss=0.05782, over 1605194.30 frames. ], batch size: 71, lr: 2.78e-03, grad_scale: 32.0 +2023-02-07 12:04:50,998 INFO [optim.py:369] (0/4) 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 12:05:10,169 WARNING [train.py:1067] (0/4) Exclude cut with ID 5239-32139-0047-92994 from training. Duration: 27.14 +2023-02-07 12:05:15,548 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215248.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 12:05:20,047 INFO [train.py:901] (0/4) Epoch 27, batch 5100, loss[loss=0.2431, simple_loss=0.2997, pruned_loss=0.0932, over 7808.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2824, pruned_loss=0.05869, over 1601394.67 frames. ], batch size: 20, lr: 2.78e-03, grad_scale: 32.0 +2023-02-07 12:05:28,227 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7796, 1.7511, 2.3482, 1.6108, 1.3875, 2.2603, 0.4565, 1.4723], + device='cuda:0'), covar=tensor([0.1879, 0.1180, 0.0316, 0.0982, 0.2567, 0.0507, 0.2129, 0.1403], + device='cuda:0'), in_proj_covar=tensor([0.0201, 0.0206, 0.0137, 0.0225, 0.0279, 0.0146, 0.0174, 0.0201], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 12:05:32,898 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.2108, 1.0649, 1.3222, 1.0447, 0.9651, 1.3010, 0.1272, 1.0261], + device='cuda:0'), covar=tensor([0.1561, 0.1417, 0.0525, 0.0660, 0.2336, 0.0610, 0.2058, 0.1265], + device='cuda:0'), in_proj_covar=tensor([0.0201, 0.0207, 0.0137, 0.0225, 0.0279, 0.0147, 0.0174, 0.0201], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 12:05:54,329 INFO [train.py:901] (0/4) Epoch 27, batch 5150, loss[loss=0.2052, simple_loss=0.3022, pruned_loss=0.05411, over 8606.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2824, pruned_loss=0.05834, over 1602057.00 frames. ], batch size: 39, lr: 2.78e-03, grad_scale: 32.0 +2023-02-07 12:05:59,266 INFO [optim.py:369] (0/4) 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 12:06:22,719 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215343.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 12:06:28,172 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215351.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:06:30,710 INFO [train.py:901] (0/4) Epoch 27, batch 5200, loss[loss=0.2138, simple_loss=0.3045, pruned_loss=0.06158, over 7977.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2823, pruned_loss=0.05772, over 1607637.42 frames. ], batch size: 21, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:07:05,197 INFO [train.py:901] (0/4) Epoch 27, batch 5250, loss[loss=0.1714, simple_loss=0.2583, pruned_loss=0.04222, over 8248.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2838, pruned_loss=0.05893, over 1607121.53 frames. ], batch size: 22, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:07:09,825 INFO [optim.py:369] (0/4) 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 12:07:12,598 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 27, batch 5300, loss[loss=0.2628, simple_loss=0.3346, pruned_loss=0.09548, over 6961.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2835, pruned_loss=0.05888, over 1606313.05 frames. ], batch size: 71, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:08:13,771 INFO [train.py:901] (0/4) Epoch 27, batch 5350, loss[loss=0.2147, simple_loss=0.291, pruned_loss=0.06921, over 7304.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2828, pruned_loss=0.05851, over 1608165.34 frames. ], batch size: 71, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:08:18,661 INFO [optim.py:369] (0/4) 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 12:08:25,712 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215521.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:08:48,872 INFO [train.py:901] (0/4) Epoch 27, batch 5400, loss[loss=0.2286, simple_loss=0.2999, pruned_loss=0.07869, over 7128.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2822, pruned_loss=0.05857, over 1605646.21 frames. ], batch size: 72, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:08:52,411 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5992, 2.5156, 1.8481, 2.3445, 2.1529, 1.5528, 2.0664, 2.1137], + device='cuda:0'), covar=tensor([0.1447, 0.0450, 0.1257, 0.0581, 0.0756, 0.1624, 0.1015, 0.0964], + device='cuda:0'), in_proj_covar=tensor([0.0361, 0.0241, 0.0343, 0.0312, 0.0303, 0.0347, 0.0350, 0.0323], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 12:08:57,717 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215566.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 12:09:08,900 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=215592.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 12:09:23,334 INFO [train.py:901] (0/4) Epoch 27, batch 5450, loss[loss=0.2197, simple_loss=0.2952, pruned_loss=0.07214, over 8458.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2822, pruned_loss=0.05847, over 1605216.52 frames. ], batch size: 25, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:09:27,892 INFO [optim.py:369] (0/4) 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] (0/4) attn_weights_entropy = tensor([2.8774, 1.9824, 2.1020, 1.9293, 1.2104, 1.9983, 2.5622, 2.5295], + device='cuda:0'), covar=tensor([0.0464, 0.1112, 0.1500, 0.1309, 0.0544, 0.1257, 0.0543, 0.0515], + device='cuda:0'), in_proj_covar=tensor([0.0099, 0.0153, 0.0189, 0.0161, 0.0101, 0.0162, 0.0112, 0.0145], + device='cuda:0'), out_proj_covar=tensor([0.0007, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:0') +2023-02-07 12:09:56,869 INFO [train.py:901] (0/4) Epoch 27, batch 5500, loss[loss=0.1855, simple_loss=0.2651, pruned_loss=0.05295, over 7915.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2824, pruned_loss=0.05866, over 1607596.19 frames. ], batch size: 20, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:09:56,883 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0014-44390_sp0.9 from training. Duration: 31.02225 +2023-02-07 12:10:20,594 INFO [zipformer.py:1185] (0/4) 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] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=215695.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:10:32,534 INFO [train.py:901] (0/4) Epoch 27, batch 5550, loss[loss=0.2046, simple_loss=0.296, pruned_loss=0.05664, over 8187.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2813, pruned_loss=0.05798, over 1607200.10 frames. ], batch size: 23, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:10:34,093 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=215707.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 12:10:37,234 INFO [optim.py:369] (0/4) 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 12:11:06,798 INFO [train.py:901] (0/4) Epoch 27, batch 5600, loss[loss=0.2099, simple_loss=0.2961, pruned_loss=0.06188, over 8523.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2816, pruned_loss=0.05793, over 1608332.71 frames. ], batch size: 26, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:11:40,020 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=215802.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 12:11:41,821 INFO [train.py:901] (0/4) Epoch 27, batch 5650, loss[loss=0.2033, simple_loss=0.2866, pruned_loss=0.06001, over 8240.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2811, pruned_loss=0.05787, over 1606766.80 frames. ], batch size: 24, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:11:46,130 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=215810.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:11:47,203 INFO [optim.py:369] (0/4) 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 12:11:50,149 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.9083, 2.3074, 3.6973, 1.9216, 1.8152, 3.6292, 0.7672, 2.2460], + device='cuda:0'), covar=tensor([0.1376, 0.1106, 0.0206, 0.1449, 0.2410, 0.0255, 0.1993, 0.1249], + device='cuda:0'), in_proj_covar=tensor([0.0199, 0.0204, 0.0135, 0.0222, 0.0275, 0.0145, 0.0171, 0.0198], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 12:12:04,983 WARNING [train.py:1067] (0/4) Exclude cut with ID 6758-72288-0033-148662_sp0.9 from training. Duration: 25.988875 +2023-02-07 12:12:11,241 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215847.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:12:16,656 INFO [train.py:901] (0/4) Epoch 27, batch 5700, loss[loss=0.2128, simple_loss=0.282, pruned_loss=0.07174, over 7413.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2808, pruned_loss=0.0573, over 1607937.77 frames. ], batch size: 17, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:12:23,686 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=215865.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:12:51,895 INFO [train.py:901] (0/4) Epoch 27, batch 5750, loss[loss=0.1588, simple_loss=0.2357, pruned_loss=0.04092, over 7647.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2804, pruned_loss=0.05685, over 1608771.39 frames. ], batch size: 19, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:12:55,997 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=215910.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 12:12:57,181 INFO [optim.py:369] (0/4) 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] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=215927.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:13:10,335 WARNING [train.py:1067] (0/4) Exclude cut with ID 3972-170212-0014-103914_sp0.9 from training. Duration: 29.1166875 +2023-02-07 12:13:26,395 INFO [train.py:901] (0/4) Epoch 27, batch 5800, loss[loss=0.2092, simple_loss=0.293, pruned_loss=0.06271, over 8195.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2824, pruned_loss=0.05794, over 1610594.21 frames. ], batch size: 23, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:13:31,949 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=215963.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 12:13:43,165 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=215980.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:13:47,578 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9048, 1.4114, 1.5389, 1.1964, 0.9375, 1.3788, 1.6273, 1.5939], + device='cuda:0'), covar=tensor([0.0527, 0.1262, 0.1777, 0.1580, 0.0608, 0.1546, 0.0710, 0.0636], + device='cuda:0'), in_proj_covar=tensor([0.0099, 0.0153, 0.0189, 0.0161, 0.0101, 0.0163, 0.0112, 0.0145], + device='cuda:0'), out_proj_covar=tensor([0.0007, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:0') +2023-02-07 12:13:48,224 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=215988.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 12:13:56,686 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-216000.pt +2023-02-07 12:14:01,060 INFO [train.py:901] (0/4) Epoch 27, batch 5850, loss[loss=0.1956, simple_loss=0.2692, pruned_loss=0.06104, over 7786.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2827, pruned_loss=0.05854, over 1609291.30 frames. ], batch size: 19, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:14:05,651 INFO [optim.py:369] (0/4) 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 12:14:15,359 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=216025.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 12:14:27,440 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=216042.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:14:30,147 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=216046.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:14:36,736 INFO [train.py:901] (0/4) Epoch 27, batch 5900, loss[loss=0.178, simple_loss=0.2628, pruned_loss=0.04657, over 8119.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2821, pruned_loss=0.0582, over 1610723.64 frames. ], batch size: 22, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:14:38,985 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=216058.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 12:14:44,134 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=216066.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:14:55,476 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=216083.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 12:15:00,975 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=216091.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:15:10,157 INFO [train.py:901] (0/4) Epoch 27, batch 5950, loss[loss=0.1784, simple_loss=0.2765, pruned_loss=0.04013, over 8104.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2823, pruned_loss=0.05777, over 1611717.40 frames. ], batch size: 23, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:15:15,769 INFO [optim.py:369] (0/4) 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 12:15:46,891 INFO [train.py:901] (0/4) Epoch 27, batch 6000, loss[loss=0.1704, simple_loss=0.2674, pruned_loss=0.03674, over 8318.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2831, pruned_loss=0.0583, over 1615871.51 frames. ], batch size: 25, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:15:46,892 INFO [train.py:926] (0/4) Computing validation loss +2023-02-07 12:15:59,961 INFO [train.py:935] (0/4) Epoch 27, validation: loss=0.1711, simple_loss=0.2711, pruned_loss=0.03554, over 944034.00 frames. +2023-02-07 12:15:59,962 INFO [train.py:936] (0/4) Maximum memory allocated so far is 6760MB +2023-02-07 12:16:25,738 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 27, batch 6050, loss[loss=0.1863, simple_loss=0.2795, pruned_loss=0.04657, over 8465.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2819, pruned_loss=0.05757, over 1617443.55 frames. ], batch size: 27, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:16:40,125 INFO [optim.py:369] (0/4) 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 12:16:56,663 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=216236.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:16:58,608 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.5744, 1.3556, 1.7241, 1.2652, 1.0016, 1.4504, 2.1429, 2.0062], + device='cuda:0'), covar=tensor([0.0505, 0.1759, 0.2385, 0.1957, 0.0676, 0.2031, 0.0745, 0.0683], + device='cuda:0'), in_proj_covar=tensor([0.0099, 0.0152, 0.0189, 0.0161, 0.0101, 0.0162, 0.0112, 0.0145], + device='cuda:0'), out_proj_covar=tensor([0.0007, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:0') +2023-02-07 12:17:09,870 INFO [train.py:901] (0/4) Epoch 27, batch 6100, loss[loss=0.1802, simple_loss=0.2545, pruned_loss=0.05298, over 7693.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2818, pruned_loss=0.05779, over 1614630.79 frames. ], batch size: 18, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:17:13,990 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=216261.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:17:23,687 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.3525, 1.8683, 3.6173, 2.0286, 2.9802, 2.8896, 3.2720, 3.2544], + device='cuda:0'), covar=tensor([0.1398, 0.4819, 0.1611, 0.4870, 0.1832, 0.1731, 0.1129, 0.1044], + device='cuda:0'), in_proj_covar=tensor([0.0672, 0.0664, 0.0734, 0.0654, 0.0744, 0.0632, 0.0634, 0.0713], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 12:17:28,455 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=216281.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 12:17:39,618 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=216298.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:17:44,969 INFO [train.py:901] (0/4) Epoch 27, batch 6150, loss[loss=0.1842, simple_loss=0.2816, pruned_loss=0.04341, over 8127.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.282, pruned_loss=0.05791, over 1614569.37 frames. ], batch size: 22, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:17:44,980 WARNING [train.py:1067] (0/4) Exclude cut with ID 3033-130750-0096-107983_sp0.9 from training. Duration: 0.92225 +2023-02-07 12:17:45,826 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=216306.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:17:45,848 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 12:17:57,198 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=216323.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:18:08,590 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.5424, 2.8450, 2.5880, 4.1305, 1.7956, 2.0342, 2.6935, 2.9197], + device='cuda:0'), covar=tensor([0.0723, 0.0769, 0.0787, 0.0206, 0.0992, 0.1181, 0.0767, 0.0811], + device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0194, 0.0244, 0.0212, 0.0203, 0.0245, 0.0249, 0.0204], + device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0006, 0.0005, 0.0005, 0.0005, 0.0006, 0.0005], + device='cuda:0') +2023-02-07 12:18:10,060 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-07 12:18:18,516 INFO [train.py:901] (0/4) Epoch 27, batch 6200, loss[loss=0.1999, simple_loss=0.2847, pruned_loss=0.05752, over 8022.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2804, pruned_loss=0.05659, over 1613939.62 frames. ], batch size: 22, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:18:42,449 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=216390.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:18:48,840 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-07 12:18:53,890 INFO [train.py:901] (0/4) Epoch 27, batch 6250, loss[loss=0.1936, simple_loss=0.2872, pruned_loss=0.04999, over 8235.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.28, pruned_loss=0.05656, over 1613028.18 frames. ], batch size: 24, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:18:58,454 INFO [optim.py:369] (0/4) 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 12:19:27,757 INFO [train.py:901] (0/4) Epoch 27, batch 6300, loss[loss=0.1795, simple_loss=0.2691, pruned_loss=0.04498, over 8187.00 frames. ], tot_loss[loss=0.1961, simple_loss=0.2798, pruned_loss=0.05624, over 1616745.21 frames. ], batch size: 23, lr: 2.78e-03, grad_scale: 16.0 +2023-02-07 12:20:01,018 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 +2023-02-07 12:20:01,970 INFO [train.py:901] (0/4) Epoch 27, batch 6350, loss[loss=0.1578, simple_loss=0.2337, pruned_loss=0.04095, over 7409.00 frames. ], tot_loss[loss=0.1959, simple_loss=0.2792, pruned_loss=0.05629, over 1610692.67 frames. ], batch size: 17, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:20:02,161 INFO [zipformer.py:1185] (0/4) 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] (0/4) 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 12:20:11,421 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.7434, 1.5268, 4.9228, 1.8522, 4.4236, 4.0722, 4.4513, 4.3679], + device='cuda:0'), covar=tensor([0.0532, 0.4755, 0.0466, 0.4273, 0.0939, 0.0972, 0.0562, 0.0563], + device='cuda:0'), in_proj_covar=tensor([0.0672, 0.0665, 0.0734, 0.0654, 0.0744, 0.0632, 0.0634, 0.0714], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 12:20:36,661 INFO [train.py:901] (0/4) Epoch 27, batch 6400, loss[loss=0.2001, simple_loss=0.2761, pruned_loss=0.06207, over 8438.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2809, pruned_loss=0.05738, over 1610004.04 frames. ], batch size: 29, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:20:41,576 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=216562.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:20:58,189 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=216587.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:21:10,659 INFO [train.py:901] (0/4) Epoch 27, batch 6450, loss[loss=0.2084, simple_loss=0.3017, pruned_loss=0.05754, over 8203.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2816, pruned_loss=0.05764, over 1614089.76 frames. ], batch size: 23, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:21:13,564 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=216609.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:21:14,324 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6453, 1.8911, 2.8185, 1.4924, 2.2470, 1.9774, 1.7355, 2.2391], + device='cuda:0'), covar=tensor([0.2136, 0.2977, 0.1041, 0.5146, 0.2029, 0.3516, 0.2711, 0.2581], + device='cuda:0'), in_proj_covar=tensor([0.0534, 0.0629, 0.0560, 0.0664, 0.0655, 0.0606, 0.0560, 0.0640], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 12:21:16,164 INFO [optim.py:369] (0/4) 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 12:21:46,015 INFO [train.py:901] (0/4) Epoch 27, batch 6500, loss[loss=0.1996, simple_loss=0.2762, pruned_loss=0.06152, over 8493.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2816, pruned_loss=0.05785, over 1614885.14 frames. ], batch size: 28, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:21:58,501 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=216673.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 12:22:04,716 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7288, 1.6869, 2.1062, 1.4836, 1.3790, 2.0888, 0.3971, 1.4128], + device='cuda:0'), covar=tensor([0.1547, 0.1143, 0.0428, 0.0869, 0.2336, 0.0439, 0.1746, 0.1152], + device='cuda:0'), in_proj_covar=tensor([0.0200, 0.0205, 0.0136, 0.0222, 0.0275, 0.0146, 0.0172, 0.0199], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 12:22:19,963 INFO [train.py:901] (0/4) Epoch 27, batch 6550, loss[loss=0.1978, simple_loss=0.2917, pruned_loss=0.05199, over 8246.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2817, pruned_loss=0.05825, over 1613746.02 frames. ], batch size: 24, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:22:25,109 INFO [optim.py:369] (0/4) 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 12:22:32,725 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=216723.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:22:48,681 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.2063, 2.0693, 2.7464, 2.2429, 2.7477, 2.2879, 2.1322, 1.6772], + device='cuda:0'), covar=tensor([0.5816, 0.5173, 0.1978, 0.4170, 0.2587, 0.3130, 0.1948, 0.5666], + device='cuda:0'), in_proj_covar=tensor([0.0969, 0.1024, 0.0834, 0.0995, 0.1032, 0.0933, 0.0775, 0.0855], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 12:22:54,586 WARNING [train.py:1067] (0/4) Exclude cut with ID 3557-8342-0013-71585 from training. Duration: 0.92 +2023-02-07 12:22:55,724 INFO [train.py:901] (0/4) Epoch 27, batch 6600, loss[loss=0.2045, simple_loss=0.2866, pruned_loss=0.06125, over 8018.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.282, pruned_loss=0.05827, over 1615080.59 frames. ], batch size: 22, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:22:59,998 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=216761.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:23:12,957 WARNING [train.py:1067] (0/4) Exclude cut with ID 4133-6541-0027-26893_sp1.1 from training. Duration: 0.9681875 +2023-02-07 12:23:17,013 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=216786.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:23:29,548 INFO [train.py:901] (0/4) Epoch 27, batch 6650, loss[loss=0.165, simple_loss=0.2457, pruned_loss=0.04213, over 7542.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2827, pruned_loss=0.05867, over 1613246.70 frames. ], batch size: 18, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:23:34,796 INFO [optim.py:369] (0/4) 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 12:24:03,764 INFO [train.py:901] (0/4) Epoch 27, batch 6700, loss[loss=0.2413, simple_loss=0.3235, pruned_loss=0.07958, over 8352.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2825, pruned_loss=0.05849, over 1615569.48 frames. ], batch size: 26, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:24:08,628 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.4450, 2.4247, 3.0611, 2.4946, 3.0890, 2.5361, 2.4570, 2.0012], + device='cuda:0'), covar=tensor([0.5920, 0.5236, 0.2309, 0.4342, 0.2988, 0.3155, 0.1933, 0.5999], + device='cuda:0'), in_proj_covar=tensor([0.0967, 0.1022, 0.0834, 0.0993, 0.1031, 0.0931, 0.0773, 0.0854], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 12:24:38,616 INFO [train.py:901] (0/4) Epoch 27, batch 6750, loss[loss=0.1884, simple_loss=0.269, pruned_loss=0.0539, over 7651.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2816, pruned_loss=0.05771, over 1614236.82 frames. ], batch size: 19, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:24:43,890 INFO [optim.py:369] (0/4) 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 12:24:56,979 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 +2023-02-07 12:25:11,160 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=216953.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:25:12,347 INFO [train.py:901] (0/4) Epoch 27, batch 6800, loss[loss=0.1782, simple_loss=0.2694, pruned_loss=0.04346, over 7980.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2824, pruned_loss=0.05818, over 1610888.85 frames. ], batch size: 21, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:25:24,102 WARNING [train.py:1067] (0/4) Exclude cut with ID 8291-282929-0024-9607_sp0.9 from training. Duration: 26.438875 +2023-02-07 12:25:32,310 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=216983.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:25:39,478 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1006, 1.7275, 3.2131, 1.7285, 2.5488, 3.5418, 3.6170, 2.9862], + device='cuda:0'), covar=tensor([0.1196, 0.1888, 0.0470, 0.2040, 0.1259, 0.0314, 0.0686, 0.0663], + device='cuda:0'), in_proj_covar=tensor([0.0305, 0.0324, 0.0292, 0.0318, 0.0321, 0.0275, 0.0438, 0.0305], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 12:25:41,670 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.71 vs. limit=5.0 +2023-02-07 12:25:43,596 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.1785, 1.3894, 4.2719, 1.9187, 2.4357, 4.8189, 4.9629, 4.1777], + device='cuda:0'), covar=tensor([0.1267, 0.2193, 0.0287, 0.2057, 0.1298, 0.0221, 0.0523, 0.0600], + device='cuda:0'), in_proj_covar=tensor([0.0305, 0.0325, 0.0292, 0.0319, 0.0321, 0.0276, 0.0438, 0.0306], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 12:25:47,365 INFO [train.py:901] (0/4) Epoch 27, batch 6850, loss[loss=0.2442, simple_loss=0.334, pruned_loss=0.07721, over 8043.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2844, pruned_loss=0.05942, over 1613794.67 frames. ], batch size: 22, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:25:52,582 INFO [optim.py:369] (0/4) 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 12:25:55,360 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=217017.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 12:26:10,983 WARNING [train.py:1067] (0/4) Exclude cut with ID 2929-85685-0079-61403_sp1.1 from training. Duration: 27.0318125 +2023-02-07 12:26:21,095 INFO [train.py:901] (0/4) Epoch 27, batch 6900, loss[loss=0.1972, simple_loss=0.2918, pruned_loss=0.05131, over 8299.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2847, pruned_loss=0.05896, over 1620994.57 frames. ], batch size: 23, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:26:29,745 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=217067.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:26:30,512 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=217068.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:26:33,610 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 +2023-02-07 12:26:56,824 INFO [train.py:901] (0/4) Epoch 27, batch 6950, loss[loss=0.2038, simple_loss=0.2903, pruned_loss=0.05864, over 8502.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2853, pruned_loss=0.05926, over 1618640.14 frames. ], batch size: 28, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:27:02,039 INFO [optim.py:369] (0/4) 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 12:27:15,666 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=217132.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 12:27:19,478 WARNING [train.py:1067] (0/4) Exclude cut with ID 7255-291500-0009-9471_sp0.9 from training. Duration: 26.62775 +2023-02-07 12:27:23,884 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 +2023-02-07 12:27:30,834 INFO [train.py:901] (0/4) Epoch 27, batch 7000, loss[loss=0.2248, simple_loss=0.31, pruned_loss=0.06978, over 8465.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2848, pruned_loss=0.05895, over 1613040.34 frames. ], batch size: 29, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:27:46,174 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.5217, 1.7887, 4.5902, 2.2850, 2.7673, 5.1461, 5.2662, 4.3225], + device='cuda:0'), covar=tensor([0.1190, 0.1921, 0.0272, 0.1935, 0.1115, 0.0254, 0.0654, 0.0662], + device='cuda:0'), in_proj_covar=tensor([0.0305, 0.0324, 0.0292, 0.0318, 0.0321, 0.0276, 0.0438, 0.0306], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 12:27:49,476 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=217182.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:28:05,134 INFO [train.py:901] (0/4) Epoch 27, batch 7050, loss[loss=0.2177, simple_loss=0.2923, pruned_loss=0.0716, over 8366.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2843, pruned_loss=0.05884, over 1609526.38 frames. ], batch size: 24, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:28:11,286 INFO [optim.py:369] (0/4) 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 12:28:40,063 INFO [train.py:901] (0/4) Epoch 27, batch 7100, loss[loss=0.1947, simple_loss=0.275, pruned_loss=0.05721, over 7649.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2839, pruned_loss=0.05888, over 1605833.20 frames. ], batch size: 19, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:29:14,492 INFO [train.py:901] (0/4) Epoch 27, batch 7150, loss[loss=0.1746, simple_loss=0.2571, pruned_loss=0.04609, over 7555.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2839, pruned_loss=0.05884, over 1603641.45 frames. ], batch size: 18, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:29:14,869 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 +2023-02-07 12:29:16,708 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7910, 1.6507, 2.1276, 1.8740, 1.8961, 1.8622, 1.6276, 0.9387], + device='cuda:0'), covar=tensor([0.6320, 0.5033, 0.1928, 0.3200, 0.2470, 0.3884, 0.2774, 0.4345], + device='cuda:0'), in_proj_covar=tensor([0.0965, 0.1023, 0.0833, 0.0992, 0.1030, 0.0930, 0.0771, 0.0852], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 12:29:19,666 INFO [optim.py:369] (0/4) 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 12:29:27,197 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6119, 2.0274, 2.9577, 1.5498, 2.2941, 2.0681, 1.7164, 2.3029], + device='cuda:0'), covar=tensor([0.1942, 0.2545, 0.0875, 0.4692, 0.1800, 0.3172, 0.2404, 0.2161], + device='cuda:0'), in_proj_covar=tensor([0.0541, 0.0634, 0.0565, 0.0669, 0.0659, 0.0612, 0.0565, 0.0645], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 12:29:27,885 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=217324.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:29:29,687 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=217327.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:29:39,080 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7894, 1.6717, 2.5322, 1.9980, 2.2773, 1.8482, 1.6099, 1.2011], + device='cuda:0'), covar=tensor([0.7960, 0.6559, 0.2596, 0.4743, 0.3715, 0.4848, 0.3234, 0.6385], + device='cuda:0'), in_proj_covar=tensor([0.0963, 0.1020, 0.0831, 0.0991, 0.1028, 0.0927, 0.0769, 0.0850], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], + device='cuda:0') +2023-02-07 12:29:45,927 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=217349.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:29:49,820 INFO [train.py:901] (0/4) Epoch 27, batch 7200, loss[loss=0.2457, simple_loss=0.3101, pruned_loss=0.09061, over 7931.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2842, pruned_loss=0.0595, over 1604731.49 frames. ], batch size: 20, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:30:11,967 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=217388.0, num_to_drop=1, layers_to_drop={1} +2023-02-07 12:30:23,036 INFO [train.py:901] (0/4) Epoch 27, batch 7250, loss[loss=0.2006, simple_loss=0.2858, pruned_loss=0.0577, over 8034.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2843, pruned_loss=0.05929, over 1609002.82 frames. ], batch size: 22, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:30:23,856 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=217406.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:30:28,398 INFO [optim.py:369] (0/4) 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 12:30:28,631 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=217413.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 12:30:45,938 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=217438.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:30:49,264 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=217442.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:30:58,440 INFO [train.py:901] (0/4) Epoch 27, batch 7300, loss[loss=0.1879, simple_loss=0.2778, pruned_loss=0.04902, over 7976.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2831, pruned_loss=0.05867, over 1606455.01 frames. ], batch size: 21, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:31:04,049 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=217463.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:31:33,092 INFO [train.py:901] (0/4) Epoch 27, batch 7350, loss[loss=0.1897, simple_loss=0.283, pruned_loss=0.0482, over 8042.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2829, pruned_loss=0.05853, over 1606856.11 frames. ], batch size: 22, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:31:36,614 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=217510.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:31:37,298 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6108, 1.4579, 1.7120, 1.3989, 0.9699, 1.5050, 1.4906, 1.0970], + device='cuda:0'), covar=tensor([0.0609, 0.1209, 0.1585, 0.1474, 0.0582, 0.1452, 0.0717, 0.0742], + device='cuda:0'), in_proj_covar=tensor([0.0099, 0.0153, 0.0189, 0.0162, 0.0101, 0.0163, 0.0112, 0.0146], + device='cuda:0'), out_proj_covar=tensor([0.0007, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:0') +2023-02-07 12:31:38,492 INFO [optim.py:369] (0/4) 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 12:31:59,834 WARNING [train.py:1067] (0/4) Exclude cut with ID 7357-94126-0026-44402_sp0.9 from training. Duration: 25.061125 +2023-02-07 12:32:03,454 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.0708, 1.6000, 1.3183, 1.6048, 1.2820, 1.1727, 1.3514, 1.4203], + device='cuda:0'), covar=tensor([0.1126, 0.0554, 0.1578, 0.0533, 0.0886, 0.1762, 0.0971, 0.0807], + device='cuda:0'), in_proj_covar=tensor([0.0363, 0.0243, 0.0345, 0.0315, 0.0306, 0.0350, 0.0353, 0.0327], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 12:32:07,059 INFO [train.py:901] (0/4) Epoch 27, batch 7400, loss[loss=0.2434, simple_loss=0.3201, pruned_loss=0.08333, over 8656.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2836, pruned_loss=0.05866, over 1610283.03 frames. ], batch size: 49, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:32:19,117 WARNING [train.py:1067] (0/4) Exclude cut with ID 774-127930-0014-48411_sp1.1 from training. Duration: 0.95 +2023-02-07 12:32:23,935 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.0296, 1.8532, 2.2013, 1.8136, 1.0448, 1.9161, 2.2728, 2.3156], + device='cuda:0'), covar=tensor([0.0448, 0.1147, 0.1461, 0.1302, 0.0551, 0.1290, 0.0599, 0.0552], + device='cuda:0'), in_proj_covar=tensor([0.0099, 0.0152, 0.0189, 0.0161, 0.0100, 0.0163, 0.0112, 0.0145], + device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:0') +2023-02-07 12:32:28,668 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.5950, 1.8151, 2.6522, 1.5296, 2.1591, 1.9111, 1.6898, 2.1986], + device='cuda:0'), covar=tensor([0.1775, 0.2432, 0.0761, 0.4126, 0.1577, 0.2769, 0.2105, 0.2044], + device='cuda:0'), in_proj_covar=tensor([0.0544, 0.0637, 0.0567, 0.0672, 0.0661, 0.0614, 0.0567, 0.0647], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 12:32:40,306 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=2.71 vs. limit=5.0 +2023-02-07 12:32:42,453 INFO [train.py:901] (0/4) Epoch 27, batch 7450, loss[loss=0.1829, simple_loss=0.2597, pruned_loss=0.05305, over 7187.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2844, pruned_loss=0.05924, over 1605930.18 frames. ], batch size: 16, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:32:47,769 INFO [optim.py:369] (0/4) 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] (0/4) Exclude cut with ID 7699-105389-0094-102071_sp0.9 from training. Duration: 26.6166875 +2023-02-07 12:33:16,132 INFO [train.py:901] (0/4) Epoch 27, batch 7500, loss[loss=0.2502, simple_loss=0.3153, pruned_loss=0.09251, over 8337.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2841, pruned_loss=0.05929, over 1604153.04 frames. ], batch size: 25, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:33:28,161 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 +2023-02-07 12:33:34,961 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=217682.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:33:46,528 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=217698.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:33:51,445 INFO [train.py:901] (0/4) Epoch 27, batch 7550, loss[loss=0.2187, simple_loss=0.2934, pruned_loss=0.07197, over 7698.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2836, pruned_loss=0.05947, over 1601470.49 frames. ], batch size: 18, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:33:56,745 INFO [optim.py:369] (0/4) 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 12:34:01,687 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.3183, 1.1863, 2.3830, 1.3587, 2.1707, 2.4897, 2.7009, 2.1379], + device='cuda:0'), covar=tensor([0.1315, 0.1614, 0.0433, 0.2033, 0.0790, 0.0453, 0.0866, 0.0698], + device='cuda:0'), in_proj_covar=tensor([0.0305, 0.0322, 0.0291, 0.0317, 0.0319, 0.0276, 0.0437, 0.0305], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 12:34:03,655 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=217723.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:34:09,252 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 +2023-02-07 12:34:21,815 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=217750.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:34:25,179 INFO [train.py:901] (0/4) Epoch 27, batch 7600, loss[loss=0.1901, simple_loss=0.2784, pruned_loss=0.05091, over 7980.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2836, pruned_loss=0.05901, over 1603010.47 frames. ], batch size: 21, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:34:53,128 INFO [scaling.py:679] (0/4) Whitening: num_groups=1, num_channels=256, metric=3.63 vs. limit=5.0 +2023-02-07 12:35:01,492 INFO [train.py:901] (0/4) Epoch 27, batch 7650, loss[loss=0.2022, simple_loss=0.2971, pruned_loss=0.05365, over 8496.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2831, pruned_loss=0.05866, over 1604760.66 frames. ], batch size: 26, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:35:06,791 INFO [optim.py:369] (0/4) 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 12:35:18,899 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([3.2692, 3.1170, 2.9139, 1.5350, 2.8240, 2.9277, 2.8447, 2.7802], + device='cuda:0'), covar=tensor([0.1096, 0.0842, 0.1240, 0.4583, 0.1160, 0.1273, 0.1597, 0.1135], + device='cuda:0'), in_proj_covar=tensor([0.0539, 0.0462, 0.0447, 0.0556, 0.0442, 0.0464, 0.0440, 0.0406], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 12:35:35,075 INFO [zipformer.py:1185] (0/4) 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] (0/4) Epoch 27, batch 7700, loss[loss=0.1847, simple_loss=0.2684, pruned_loss=0.05051, over 7816.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2832, pruned_loss=0.05828, over 1606154.79 frames. ], batch size: 20, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:35:42,366 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=217865.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:36:05,138 WARNING [train.py:1067] (0/4) 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] (0/4) Epoch 27, batch 7750, loss[loss=0.226, simple_loss=0.3036, pruned_loss=0.07417, over 8140.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2822, pruned_loss=0.05786, over 1608002.13 frames. ], batch size: 22, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:36:14,308 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 +2023-02-07 12:36:15,951 INFO [optim.py:369] (0/4) 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 12:36:18,066 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=217916.0, num_to_drop=1, layers_to_drop={0} +2023-02-07 12:36:27,433 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.7606, 1.5151, 3.2819, 1.3981, 2.4975, 3.5789, 3.7994, 2.9240], + device='cuda:0'), covar=tensor([0.1504, 0.2076, 0.0442, 0.2395, 0.1142, 0.0349, 0.0841, 0.0829], + device='cuda:0'), in_proj_covar=tensor([0.0305, 0.0322, 0.0290, 0.0317, 0.0318, 0.0275, 0.0435, 0.0304], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 12:36:45,570 INFO [train.py:901] (0/4) Epoch 27, batch 7800, loss[loss=0.1867, simple_loss=0.269, pruned_loss=0.05219, over 8088.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2812, pruned_loss=0.05718, over 1610142.74 frames. ], batch size: 21, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:36:55,070 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=217969.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:37:04,140 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.9820, 2.8839, 2.6974, 1.4436, 2.6120, 2.6756, 2.6230, 2.6044], + device='cuda:0'), covar=tensor([0.1085, 0.0871, 0.1195, 0.4397, 0.1120, 0.1253, 0.1553, 0.1079], + device='cuda:0'), in_proj_covar=tensor([0.0538, 0.0461, 0.0446, 0.0554, 0.0440, 0.0464, 0.0438, 0.0405], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 12:37:15,325 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/checkpoint-218000.pt +2023-02-07 12:37:19,654 INFO [train.py:901] (0/4) Epoch 27, batch 7850, loss[loss=0.2364, simple_loss=0.3131, pruned_loss=0.07982, over 8351.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2817, pruned_loss=0.05775, over 1612103.15 frames. ], batch size: 49, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:37:24,955 INFO [optim.py:369] (0/4) 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 12:37:33,479 INFO [zipformer.py:1185] (0/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=218026.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:37:52,839 INFO [train.py:901] (0/4) Epoch 27, batch 7900, loss[loss=0.1873, simple_loss=0.2741, pruned_loss=0.05027, over 8079.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2808, pruned_loss=0.05718, over 1610547.37 frames. ], batch size: 21, lr: 2.77e-03, grad_scale: 8.0 +2023-02-07 12:37:52,991 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([4.6514, 1.6280, 4.8490, 1.9169, 4.3220, 4.0126, 4.4264, 4.2755], + device='cuda:0'), covar=tensor([0.0537, 0.4701, 0.0448, 0.4142, 0.0931, 0.0964, 0.0551, 0.0605], + device='cuda:0'), in_proj_covar=tensor([0.0677, 0.0663, 0.0731, 0.0653, 0.0742, 0.0631, 0.0635, 0.0717], + device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], + device='cuda:0') +2023-02-07 12:37:58,506 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 +2023-02-07 12:38:08,979 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.7362, 2.1499, 3.5857, 1.6083, 1.7337, 3.5283, 0.5899, 2.0418], + device='cuda:0'), covar=tensor([0.1551, 0.1117, 0.0216, 0.1803, 0.2360, 0.0233, 0.1932, 0.1226], + device='cuda:0'), in_proj_covar=tensor([0.0201, 0.0206, 0.0137, 0.0224, 0.0278, 0.0147, 0.0173, 0.0200], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003], + device='cuda:0') +2023-02-07 12:38:09,510 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([2.7131, 1.4893, 1.9267, 1.5507, 0.8443, 1.7034, 2.0703, 1.9459], + device='cuda:0'), covar=tensor([0.0520, 0.1355, 0.1733, 0.1565, 0.0647, 0.1503, 0.0697, 0.0640], + device='cuda:0'), in_proj_covar=tensor([0.0099, 0.0153, 0.0189, 0.0161, 0.0101, 0.0162, 0.0112, 0.0146], + device='cuda:0'), out_proj_covar=tensor([0.0007, 0.0009, 0.0010, 0.0010, 0.0006, 0.0009, 0.0007, 0.0008], + device='cuda:0') +2023-02-07 12:38:25,725 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([0.9662, 1.6578, 1.3841, 1.5105, 1.3094, 1.2963, 1.3104, 1.2991], + device='cuda:0'), covar=tensor([0.1379, 0.0534, 0.1475, 0.0692, 0.0866, 0.1695, 0.0962, 0.0953], + device='cuda:0'), in_proj_covar=tensor([0.0359, 0.0242, 0.0341, 0.0313, 0.0305, 0.0346, 0.0349, 0.0323], + device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], + device='cuda:0') +2023-02-07 12:38:26,203 INFO [train.py:901] (0/4) Epoch 27, batch 7950, loss[loss=0.2345, simple_loss=0.3148, pruned_loss=0.07711, over 8314.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2809, pruned_loss=0.05732, over 1610724.47 frames. ], batch size: 25, lr: 2.76e-03, grad_scale: 8.0 +2023-02-07 12:38:31,702 INFO [optim.py:369] (0/4) 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 12:38:35,258 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218118.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:38:37,426 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=218121.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:38:50,499 INFO [zipformer.py:1185] (0/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=218141.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:38:51,659 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.9996, 1.5295, 3.4383, 1.5493, 2.4045, 3.7599, 3.8545, 3.2488], + device='cuda:0'), covar=tensor([0.1223, 0.1853, 0.0333, 0.2140, 0.1074, 0.0227, 0.0459, 0.0554], + device='cuda:0'), in_proj_covar=tensor([0.0305, 0.0323, 0.0291, 0.0318, 0.0320, 0.0276, 0.0437, 0.0305], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 12:38:51,965 INFO [scaling.py:679] (0/4) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 +2023-02-07 12:38:53,605 INFO [zipformer.py:1185] (0/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=218146.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:38:59,466 INFO [train.py:901] (0/4) Epoch 27, batch 8000, loss[loss=0.1751, simple_loss=0.2525, pruned_loss=0.04881, over 6372.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2813, pruned_loss=0.05745, over 1613665.80 frames. ], batch size: 14, lr: 2.76e-03, grad_scale: 8.0 +2023-02-07 12:39:05,390 INFO [zipformer.py:2431] (0/4) attn_weights_entropy = tensor([1.6952, 1.4145, 3.1633, 1.3520, 2.3709, 3.3856, 3.5579, 2.9272], + device='cuda:0'), covar=tensor([0.1349, 0.1908, 0.0361, 0.2275, 0.1061, 0.0268, 0.0485, 0.0549], + device='cuda:0'), in_proj_covar=tensor([0.0305, 0.0323, 0.0291, 0.0317, 0.0320, 0.0276, 0.0437, 0.0305], + device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0004, 0.0002], + device='cuda:0') +2023-02-07 12:39:29,223 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218200.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:39:32,351 INFO [train.py:901] (0/4) Epoch 27, batch 8050, loss[loss=0.1781, simple_loss=0.2657, pruned_loss=0.04524, over 8460.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2817, pruned_loss=0.05782, over 1605336.87 frames. ], batch size: 25, lr: 2.76e-03, grad_scale: 8.0 +2023-02-07 12:39:38,065 INFO [optim.py:369] (0/4) 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 12:39:46,224 INFO [zipformer.py:1185] (0/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=218225.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:39:48,246 INFO [zipformer.py:1185] (0/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218228.0, num_to_drop=0, layers_to_drop=set() +2023-02-07 12:39:55,707 INFO [checkpoint.py:75] (0/4) Saving checkpoint to pruned_transducer_stateless7_streaming/exp/v1/epoch-27.pt